Lectures on Electromagnetic Field Theory
Weng Cho CHEW1
Fall 2020, Purdue University
1Updated: December 3, 2020
Contents
Preface xi
Acknowledgements xii
1 Introduction, Maxwell’s Equations 1 1.1 Importance of Electromagnetics ...... 1 1.1.1 A Brief History of Electromagnetics ...... 4 1.2 Maxwell’s Equations in Integral Form ...... 7 1.3 Static Electromagnetics ...... 8 1.3.1 Coulomb’s Law (Statics) ...... 8 1.3.2 Electric Field E (Statics) ...... 8 1.3.3 Gauss’s Law (Statics) ...... 11 1.3.4 Derivation of Gauss’s Law from Coulomb’s Law (Statics) ...... 12 1.4 Homework Examples ...... 13
2 Maxwell’s Equations, Differential Operator Form 17 2.1 Gauss’s Divergence Theorem ...... 17 2.1.1 Some Details ...... 19 2.1.2 Gauss’s Law in Differential Operator Form ...... 21 2.1.3 Physical Meaning of Divergence Operator ...... 21 2.2 Stokes’s Theorem ...... 22 2.2.1 Faraday’s Law in Differential Operator Form ...... 25 2.2.2 Physical Meaning of Curl Operator ...... 26 2.3 Maxwell’s Equations in Differential Operator Form ...... 26 2.4 Homework Examples ...... 27 2.5 Historical Notes ...... 28
3 Constitutive Relations, Wave Equation, Electrostatics, and Static Green’s Function 29 3.1 Simple Constitutive Relations ...... 29 3.2 Emergence of Wave Phenomenon, Triumph of Maxwell’s Equations ...... 30 3.3 Static Electromagnetics–Revisted ...... 34 3.3.1 Electrostatics ...... 34
i ii Electromagnetic Field Theory
3.3.2 Poisson’s Equation ...... 35 3.3.3 Static Green’s Function ...... 36 3.3.4 Laplace’s Equation ...... 36 3.4 Homework Examples ...... 37
4 Magnetostatics, Boundary Conditions, and Jump Conditions 39 4.1 Magnetostatics ...... 39 4.1.1 More on Coulomb Gauge ...... 41 4.2 Boundary Conditions–1D Poisson’s Equation ...... 41 4.3 Boundary Conditions–Maxwell’s Equations ...... 43 4.3.1 Faraday’s Law ...... 44 4.3.2 Gauss’s Law for Electric Flux ...... 45 4.3.3 Ampere’s Law ...... 46 4.3.4 Gauss’s Law for Magnetic Flux ...... 48
5 Biot-Savart law, Conductive Media Interface, Instantaneous Poynting’s Theorem 49 5.1 Derivation of Biot-Savart Law ...... 50 5.2 Shielding by Conductive Media ...... 52 5.2.1 Boundary Conditions–Conductive Media Case ...... 52 5.2.2 Electric Field Inside a Conductor ...... 53 5.2.3 Magnetic Field Inside a Conductor ...... 54 5.3 Instantaneous Poynting’s Theorem ...... 56
6 Time-Harmonic Fields, Complex Power 61 6.1 Time-Harmonic Fields—Linear Systems ...... 62 6.2 Fourier Transform Technique ...... 64 6.3 Complex Power ...... 65
7 More on Constitute Relations, Uniform Plane Wave 69 7.1 More on Constitutive Relations ...... 69 7.1.1 Isotropic Frequency Dispersive Media ...... 69 7.1.2 Anisotropic Media ...... 71 7.1.3 Bi-anisotropic Media ...... 72 7.1.4 Inhomogeneous Media ...... 72 7.1.5 Uniaxial and Biaxial Media ...... 72 7.1.6 Nonlinear Media ...... 73 7.2 Wave Phenomenon in the Frequency Domain ...... 73 7.3 Uniform Plane Waves in 3D ...... 75
8 Lossy Media, Lorentz Force Law, Drude-Lorentz-Sommerfeld Model 79 8.1 Plane Waves in Lossy Conductive Media ...... 79 8.1.1 Highly Conductive Case ...... 80 8.1.2 Lowly Conductive Case ...... 81 8.2 Lorentz Force Law ...... 82 8.3 Drude-Lorentz-Sommerfeld Model ...... 82 Contents iii
8.3.1 Cold Collisionless Plasma Medium ...... 83 8.3.2 Bound Electron Case ...... 85 8.3.3 Damping or Dissipation Case ...... 85 8.3.4 Broad Applicability of Drude-Lorentz-Sommerfeld Model ...... 86 8.3.5 Frequency Dispersive Media ...... 88 8.3.6 Plasmonic Nanoparticles ...... 89
9 Waves in Gyrotropic Media, Polarization 91 9.1 Gyrotropic Media and Faraday Rotation ...... 91 9.2 Wave Polarization ...... 94 9.2.1 Arbitrary Polarization Case and Axial Ratio1 ...... 96 9.3 Polarization and Power Flow ...... 98
10 Spin Angular Momentum, Complex Poynting’s Theorem, Lossless Condi- tion, Energy Density 101 10.1 Spin Angular Momentum and Cylindrical Vector Beam ...... 102 10.2 Momentum Density of Electromagnetic Field ...... 103 10.3 Complex Poynting’s Theorem and Lossless Conditions ...... 104 10.3.1 Complex Poynting’s Theorem ...... 104 10.3.2 Lossless Conditions ...... 105 10.4 Energy Density in Dispersive Media ...... 107
11 Transmission Lines 111 11.1 Transmission Line Theory ...... 112 11.1.1 Time-Domain Analysis ...... 113 11.1.2 Frequency-Domain Analysis–the Power of Phasor Technique Again! . . 116 11.2 Lossy Transmission Line ...... 117
12 More on Transmission Lines 121 12.1 Terminated Transmission Lines ...... 121 12.1.1 Shorted Terminations ...... 124 12.1.2 Open Terminations ...... 125 12.2 Smith Chart ...... 126 12.3 VSWR (Voltage Standing Wave Ratio) ...... 128
13 Multi-Junction Transmission Lines, Duality Principle 133 13.1 Multi-Junction Transmission Lines ...... 133 13.1.1 Single-Junction Transmission Lines ...... 135 13.1.2 Two-Junction Transmission Lines ...... 136 13.1.3 Stray Capacitance and Inductance ...... 139 13.2 Duality Principle ...... 140 13.2.1 Unusual Swaps2 ...... 141 13.3 Fictitious Magnetic Currents ...... 142
1This section is mathematically complicated. It can be skipped on first reading. 2This section can be skipped on first reading. iv Electromagnetic Field Theory
14 Reflection, Transmission, and Interesting Physical Phenomena 145 14.1 Reflection and Transmission—Single Interface Case ...... 145 14.1.1 TE Polarization (Perpendicular or E Polarization)3 ...... 146 14.1.2 TM Polarization (Parallel or H Polarization)4 ...... 148 14.2 Interesting Physical Phenomena ...... 149 14.2.1 Total Internal Reflection ...... 150
15 More on Interesting Physical Phenomena, Homomorphism, Plane Waves, and Transmission Lines 155 15.1 Brewster’s Angle ...... 155 15.1.1 Surface Plasmon Polariton ...... 158 15.2 Homomorphism of Uniform Plane Waves and Transmission Lines Equations . 160 15.2.1 TE or TEz Waves ...... 161 15.2.2 TM or TMz Waves ...... 162
16 Waves in Layered Media 165 16.1 Waves in Layered Media ...... 165 16.1.1 Generalized Reflection Coefficient for Layered Media ...... 166 16.1.2 Ray Series Interpretation of Generalized Reflection Coefficient . . . . 167 16.1.3 Guided Modes from Generalized Reflection Coefficients ...... 168 16.2 Phase Velocity and Group Velocity ...... 168 16.2.1 Phase Velocity ...... 168 16.2.2 Group Velocity ...... 170 16.3 Wave Guidance in a Layered Media ...... 173 16.3.1 Transverse Resonance Condition ...... 173
17 Dielectric Waveguides 175 17.1 Generalized Transverse Resonance Condition ...... 175 17.2 Dielectric Waveguide ...... 176 17.2.1 TE Case ...... 177 17.2.2 TM Case ...... 183 17.2.3 A Note on Cut-Off of Dielectric Waveguides ...... 184
18 Hollow Waveguides 185 18.1 General Information on Hollow Waveguides ...... 185 18.1.1 Absence of TEM Mode in a Hollow Waveguide ...... 186 18.1.2 TE Case (Ez = 0, Hz 6= 0, TEz case) ...... 187 18.1.3 TM Case (Ez 6= 0, Hz = 0, TMz Case) ...... 189 18.2 Rectangular Waveguides ...... 190 18.2.1 TE Modes (Hz 6= 0, H Modes or TEz Modes) ...... 190
3 These polarizations are also variously know as TEz, or the s and p polarizations, a descendent from the notations for acoustic waves where s and p stand for shear and pressure waves respectively. 4 Also known as TMz polarization. Contents v
19 More on Hollow Waveguides 193 19.1 Rectangular Waveguides, Contd...... 194 19.1.1 TM Modes (Ez 6= 0, E Modes or TMz Modes) ...... 194 19.1.2 Bouncing Wave Picture ...... 195 19.1.3 Field Plots ...... 196 19.2 Circular Waveguides ...... 198 19.2.1 TE Case ...... 198 19.2.2 TM Case ...... 200
20 More on Waveguides and Transmission Lines 205 20.1 Circular Waveguides, Contd...... 205 20.1.1 An Application of Circular Waveguide ...... 206 20.2 Remarks on Quasi-TEM Modes, Hybrid Modes, and Surface Plasmonic Modes 209 20.2.1 Quasi-TEM Modes ...... 209 20.2.2 Hybrid Modes–Inhomogeneously-Filled Waveguides ...... 210 20.2.3 Guidance of Modes ...... 211 20.3 Homomorphism of Waveguides and Transmission Lines ...... 212 20.3.1 TE Case ...... 212 20.3.2 TM Case ...... 214 20.3.3 Mode Conversion ...... 216
21 Cavity Resonators 219 21.1 Transmission Line Model of a Resonator ...... 219 21.2 Cylindrical Waveguide Resonators ...... 221 21.2.1 βz = 0 Case ...... 223 21.2.2 Lowest Mode of a Rectangular Cavity ...... 224 21.3 Some Applications of Resonators ...... 225 21.3.1 Filters ...... 226 21.3.2 Electromagnetic Sources ...... 227 21.3.3 Frequency Sensor ...... 230
22 Quality Factor of Cavities, Mode Orthogonality 233 22.1 The Quality Factor of a Cavity–General Concept ...... 233 22.1.1 Analogue with an LC Tank Circuit ...... 234 22.1.2 Relation to Bandwidth and Pole Location ...... 236 22.1.3 Wall Loss and Q for a Metallic Cavity ...... 237 22.1.4 Example: The Q of TM110 Mode ...... 239 22.2 Mode Orthogonality and Matrix Eigenvalue Problem ...... 240 22.2.1 Matrix Eigenvalue Problem (EVP) ...... 240 22.2.2 Homomorphism with the Waveguide Mode Problem ...... 241 22.2.3 Proof of Orthogonality of Waveguide Modes5 ...... 242
5This may be skipped on first reading. vi Electromagnetic Field Theory
23 Scalar and Vector Potentials 245 23.1 Scalar and Vector Potentials for Time-Harmonic Fields ...... 245 23.2 Scalar and Vector Potentials for Statics, A Review ...... 246 23.2.1 Scalar and Vector Potentials for Electrodynamics ...... 247 23.2.2 More on Scalar and Vector Potentials ...... 249 23.3 When is Static Electromagnetic Theory Valid? ...... 250 23.3.1 Quasi-Static Electromagnetic Theory ...... 255
24 Circuit Theory Revisited 257 24.1 Kirchhoff Current Law ...... 257 24.2 Kirchhoff Voltage Law ...... 258 24.3 Inductor ...... 262 24.4 Capacitance ...... 263 24.5 Resistor ...... 263 24.6 Some Remarks ...... 264 24.7 Energy Storage Method for Inductor and Capacitor ...... 264 24.8 Finding Closed-Form Formulas for Inductance and Capacitance ...... 265 24.9 Importance of Circuit Theory in IC Design ...... 267 24.10Decoupling Capacitors and Spiral Inductors ...... 269
25 Radiation by a Hertzian Dipole 271 25.1 History ...... 271 25.2 Approximation by a Point Source ...... 273 25.2.1 Case I. Near Field, βr 1 ...... 275 25.2.2 Case II. Far Field (Radiation Field), βr 1 ...... 276 25.3 Radiation, Power, and Directive Gain Patterns ...... 276 25.3.1 Radiation Resistance ...... 279
26 Radiation Fields 283 26.1 Radiation Fields or Far-Field Approximation ...... 284 26.1.1 Far-Field Approximation ...... 285 26.1.2 Locally Plane Wave Approximation ...... 286 26.1.3 Directive Gain Pattern Revisited ...... 289
27 Array Antennas, Fresnel Zone, Rayleigh Distance 293 27.1 Linear Array of Dipole Antennas ...... 293 27.1.1 Far-Field Approximation ...... 294 27.1.2 Radiation Pattern of an Array ...... 295 27.2 When is Far-Field Approximation Valid? ...... 298 27.2.1 Rayleigh Distance ...... 299 27.2.2 Near Zone, Fresnel Zone, and Far Zone ...... 300 Contents vii
28 Different Types of Antennas—Heuristics 303 28.1 Resonance Tunneling in Antenna ...... 304 28.2 Horn Antennas ...... 307 28.3 Quasi-Optical Antennas ...... 309 28.4 Small Antennas ...... 311
29 Uniqueness Theorem 317 29.1 The Difference Solutions to Source-Free Maxwell’s Equations ...... 317 29.2 Conditions for Uniqueness ...... 320 29.2.1 Isotropic Case ...... 320 29.2.2 General Anisotropic Case ...... 321 29.3 Hind Sight Using Linear Algebra ...... 322 29.4 Connection to Poles of a Linear System ...... 323 29.5 Radiation from Antenna Sources and Radiation Condition ...... 324
30 Reciprocity Theorem 327 30.1 Mathematical Derivation ...... 328 30.2 Conditions for Reciprocity ...... 331 30.3 Application to a Two-Port Network and Circuit Theory ...... 331 30.4 Voltage Sources in Electromagnetics ...... 333 30.5 Hind Sight ...... 334 30.6 Transmit and Receive Patterns of an Antennna ...... 335
31 Equivalence Theorems, Huygens’ Principle 339 31.1 Equivalence Theorems or Equivalence Principles ...... 339 31.1.1 Inside-Out Case ...... 340 31.1.2 Outside-in Case ...... 341 31.1.3 General Case ...... 341 31.2 Electric Current on a PEC ...... 342 31.3 Magnetic Current on a PMC ...... 343 31.4 Huygens’ Principle and Green’s Theorem ...... 343 31.4.1 Scalar Waves Case ...... 344 31.4.2 Electromagnetic Waves Case ...... 346
32 Shielding, Image Theory 349 32.1 Shielding ...... 349 32.1.1 A Note on Electrostatic Shielding ...... 349 32.1.2 Relaxation Time ...... 350 32.2 Image Theory ...... 351 32.2.1 Electric Charges and Electric Dipoles ...... 352 32.2.2 Magnetic Charges and Magnetic Dipoles ...... 353 32.2.3 Perfect Magnetic Conductor (PMC) Surfaces ...... 354 32.2.4 Multiple Images ...... 355 32.2.5 Some Special Cases—Spheres, Cylinders, and Dielectric Interfaces . . 356 viii Electromagnetic Field Theory
33 High Frequency Solutions, Gaussian Beams 361 33.1 Tangent Plane Approximations ...... 362 33.2 Fermat’s Principle ...... 363 33.2.1 Generalized Snell’s Law ...... 365 33.3 Gaussian Beam ...... 366 33.3.1 Derivation of the Paraxial/Parabolic Wave Equation ...... 366 33.3.2 Finding a Closed Form Solution ...... 367 33.3.3 Other solutions ...... 369
34 Scattering of Electromagnetic Field 371 34.1 Rayleigh Scattering ...... 371 34.1.1 Scattering by a Small Spherical Particle ...... 373 34.1.2 Scattering Cross Section ...... 375 34.1.3 Small Conductive Particle ...... 378 34.2 Mie Scattering ...... 379 34.2.1 Optical Theorem ...... 380 34.2.2 Mie Scattering by Spherical Harmonic Expansions ...... 381 34.2.3 Separation of Variables in Spherical Coordinates6 ...... 381
35 Spectral Expansions of Source Fields—Sommerfeld Integrals 383 35.1 Spectral Representations of Sources ...... 383 35.1.1 A Point Source ...... 384 35.2 A Source on Top of a Layered Medium ...... 389 35.2.1 Electric Dipole Fields–Spectral Expansion ...... 389 35.3 Stationary Phase Method—Fermat’s Principle ...... 392 35.4 Riemann Sheets and Branch Cuts7 ...... 396 35.5 Some Remarks8 ...... 396
36 Computational Electromagnetics, Numerical Methods 399 36.1 Computational Electromagnetics, Numerical Methods ...... 401 36.2 Examples of Differential Equations ...... 401 36.3 Examples of Integral Equations ...... 402 36.3.1 Volume Integral Equation ...... 402 36.3.2 Surface Integral Equation ...... 404 36.4 Function as a Vector ...... 405 36.5 Operator as a Map ...... 406 36.5.1 Domain and Range Spaces ...... 406 36.6 Approximating Operator Equations with Matrix Equations ...... 407 36.6.1 Subspace Projection Methods ...... 407 36.6.2 Dual Spaces ...... 408 36.6.3 Matrix and Vector Representations ...... 408 36.6.4 Mesh Generation ...... 409
6May be skipped on first reading. 7This may be skipped on first reading. 8This may be skipped on first reading. Contents ix
36.6.5 Differential Equation Solvers versus Integral Equation Solvers . . . . . 410 36.7 Solving Matrix Equation by Optimization ...... 410 36.7.1 Gradient of a Functional ...... 411
37 Finite Difference Method, Yee Algorithm 415 37.1 Finite-Difference Time-Domain Method ...... 415 37.1.1 The Finite-Difference Approximation ...... 416 37.1.2 Time Stepping or Time Marching ...... 418 37.1.3 Stability Analysis ...... 420 37.1.4 Grid-Dispersion Error ...... 422 37.2 The Yee Algorithm ...... 424 37.2.1 Finite-Difference Frequency Domain Method ...... 427 37.3 Absorbing Boundary Conditions ...... 428
38 Quantum Theory of Light 431 38.1 Historical Background on Quantum Theory ...... 431 38.2 Connecting Electromagnetic Oscillation to Simple Pendulum ...... 434 38.3 Hamiltonian Mechanics ...... 438 38.4 Schr¨odingerEquation (1925) ...... 440 38.5 Some Quantum Interpretations–A Preview ...... 443 38.5.1 Matrix or Operator Representations ...... 444 38.6 Bizarre Nature of the Photon Number States ...... 445
39 Quantum Coherent State of Light 447 39.1 The Quantum Coherent State ...... 447 39.1.1 Quantum Harmonic Oscillator Revisited ...... 448 39.2 Some Words on Quantum Randomness and Quantum Observables ...... 450 39.3 Derivation of the Coherent States ...... 451 39.3.1 Time Evolution of a Quantum State ...... 453 39.4 More on the Creation and Annihilation Operator ...... 454 39.4.1 Connecting Quantum Pendulum to Electromagnetic Oscillator9 . . . . 457 39.5 Epilogue ...... 460
9May be skipped on first reading. x Electromagnetic Field Theory Preface
This set of lecture notes is from my teaching of ECE 604, Electromagnetic Field Theory, at ECE, Purdue University, West Lafayette. It is intended for entry level graduate students. Because different universities have different undergraduate requirements in electromagnetic field theory, this is a course intended to “level the playing field”. From this point onward, hopefully, all students will have the fundamental background in electromagnetic field theory needed to take advance level courses and do research at Purdue. In developing this course, I have drawn heavily upon knowledge of our predecessors in this area. Many of the textbooks and papers used, I have listed them in the reference list. Being a practitioner in this field for over 40 years, I have seen electromagnetic theory im- pacting modern technology development unabated. Despite its age, the set of Maxwell’s equations has endured and continued to be important, from statics to optics, from classi- cal to quantum, and from nanometer lengthscales to galactic lengthscales. The applications of electromagnetic technologies have also been tremendous and wide-ranging: from geophysi- cal exploration, remote sensing, bio-sensing, electrical machinery, renewable and clean energy, biomedical engineering, optics and photonics, computer chip, computer system, and quantum computer designs, quantum communication and many more. Electromagnetic field theory is not everything, but is remains an important component of modern technology developments. The challenge in teaching this course is on how to teach over 150 years of knowledge in one semester: Of course this is mission impossible! To do this, we use the traditional wisdom of engineering education: Distill the knowledge, make it as simple as possible, and teach the fundamental big ideas in one short semester. Because of this, you may find the flow of the lectures erratic. Some times, I feel the need to touch on certain big ideas before moving on, resulting in the choppiness of the curriculum. Also, in this course, I exploit mathematical homomorphism as much as possible to simplify the teaching. After years of practising in this area, I find that some complex and advanced concepts become simpler if mathematical homomorphism is exploited between the advanced concepts and simpler ones. An example of this is on waves in layered media. The problem is homomorphic to the transmission line problem: Hence, using transmission line theory, one can simplify the derivations of some complicated formulas. A large part of modern electromagnetic technologies is based on heuristics. This is some- thing difficult to teach, as it relies on physical insight and experience. Modern commercial software has reshaped this landscape, as the field of math-physics modeling through numer- ical simulations, known as computational electromagnetic (CEM), has made rapid advances in recent years. Many cut-and-try laboratory experiments, based on heuristics, have been
xi xii Lectures on Electromagnetic Field Theory replaced by cut-and-try computer experiments, which are a lot cheaper. An exciting modern development is the role of electromagnetics and Maxwell’s equations in quantum technologies. We will connect Maxwell’s equations to them toward the end of this course. This is a challenge, as it has never been done before at an entry level course to my knowledge.
Weng Cho CHEW December 3, 2020 Purdue University
Acknowledgements I like to thank Dan Jiao for sharing her lecture notes in this course, as well as Andy Weiner for sharing his experience in teaching this course. Mark Lundstrom gave me useful feedback on Lectures 38 and 39 on the quantum theory of light. Also, I am thankful to Dong-Yeop Na for helping teach part of this course. Robert Hsueh-Yung Chao also took time to read the lecture notes and gave me very useful feedback. Lecture 1
Introduction, Maxwell’s Equations
In the beginning, this field is either known as electricity and magnetism or optics. But later, as we shall discuss, these two fields are found to be based on the same set equations known as Maxwell’s equations. Maxwell’s equations unified these two fields, and it is common to call the study of electromagnetic theory based on Maxwell’s equations electromagnetics. It has wide-ranging applications from statics to ultra-violet light in the present world with impact on many different technologies.
1.1 Importance of Electromagnetics
We will explain why electromagnetics is so important, and its impact on very many different areas. Then we will give a brief history of electromagnetics, and how it has evolved in the modern world. Next we will go briefly over Maxwell’s equations in their full glory. But we will begin the study of electromagnetics by focussing on static problems which are valid in the long-wavelength limit. It has been based on Maxwell’s equations, which are the result of the seminal work of James Clerk Maxwell completed in 1865, after his presentation to the British Royal Society in 1864. It has been over 150 years ago now, and this is a long time compared to the leaps and bounds progress we have made in technological advancements. Nevertheless, research in electromagnetics has continued unabated despite its age. The reason is that electromagnetics is extremely useful, and has impacted a large sector of modern technologies. To understand why electromagnetics is so useful, we have to understand a few points about Maxwell’s equations. Maxwell’s equations are valid over a vast length scale from subatomic dimensions to galactic dimensions. Hence, these equations are valid over a vast range of wavelengths, going from static to ultra-violet wavelengths.1
1Current lithography process is working with using ultra-violet light with a wavelength of 193 nm.
1 2 Electromagnetic Field Theory
Maxwell’s equations are relativistic invariant in the parlance of special relativity [1]. In fact, Einstein was motivated with the theory of special relativity in 1905 by Maxwell’s equations [2]. These equations look the same, irrespective of what inertial reference frame one is in.
Maxwell’s equations are valid in the quantum regime, as it was demonstrated by Paul Dirac in 1927 [3]. Hence, many methods of calculating the response of a medium to classical field can be applied in the quantum regime also. When electromagnetic theory is combined with quantum theory, the field of quantum optics came about. Roy Glauber won a Nobel prize in 2005 because of his work in this area [4].
Maxwell’s equations and the pertinent gauge theory has inspired Yang-Mills theory (1954) [5], which is also known as a generalized electromagnetic theory. Yang-Mills theory is motivated by differential forms in differential geometry [6]. To quote from Misner, Thorne, and Wheeler, “Differential forms illuminate electromagnetic theory, and electromagnetic theory illuminates differential forms.” [7, 8]
Maxwell’s equations are some of the most accurate physical equations that have been validated by experiments. In 1985, Richard Feynman wrote that electromagnetic theory had been validated to one part in a billion.2 Now, it has been validated to one part in a trillion (Aoyama et al, Styer, 2012).3
As a consequence, electromagnetics has had a tremendous impact in science and tech- nology. This is manifested in electrical engineering, optics, wireless and optical commu- nications, computers, remote sensing, bio-medical engineering etc.
2This means that if a jet is to fly from New York to Los Angeles, an error of one part in a billion means an error of a few millmeters. 3This means an error of a hairline, if one were to fly from the earth to the moon. Introduction, Maxwell’s Equations 3
Figure 1.1: The impact of electromagnetics in many technologies. The areas in blue are prevalent areas impacted by electromagnetics some 20 years ago [9], and the areas in brown are modern emerging areas impacted by electromagnetics. 4 Electromagnetic Field Theory
Figure 1.2: Knowledge grows like a tree. Engineering knowledge and real-world applica- tions are driven by fundamental knowledge from math and sciences. At a university, we do science-based engineering research that can impact wide-ranging real-world applications. But everyone is equally important in transforming our society. Just like the parts of the human body, no one can claim that one is more important than the others.
Figure 1.2 shows how knowledge are driven by basic math and science knowledge. Its growth is like a tree. It is important that we collaborate to develop technologies that can transform this world.
1.1.1 A Brief History of Electromagnetics Electricity and magnetism have been known to mankind for a long time. Also, the physical properties of light have been known. But electricity and magnetism, now termed electromag- netics in the modern world, has been thought to be governed by different physical laws as opposed to optics. This is understandable as the physics of electricity and magnetism is quite different of the physics of optics as they were known to humans then. For example, lode stone was known to the ancient Greek and Chinese around 600 BC to 400 BC. Compass was used in China since 200 BC. Static electricity was reported by the Greek as early as 400 BC. But these curiosities did not make an impact until the age of telegraphy. The coming about of telegraphy was due to the invention of the voltaic cell or the galvanic cell in the late 1700’s, by Luigi Galvani and Alesandro Volta [10]. It was soon discovered that two pieces of wire, connected to a voltaic cell, can transmit information. Introduction, Maxwell’s Equations 5
So by the early 1800’s this possibility had spurred the development of telegraphy. Both Andr´e-MarieAmp´ere(1823) [11, 12] and Michael Faraday (1838) [13] did experiments to better understand the properties of electricity and magnetism. And hence, Ampere’s law and Faraday law are named after them. Kirchhoff voltage and current laws were also developed in 1845 to help better understand telegraphy [14, 15]. Despite these laws, the technology of telegraphy was poorly understood. For instance, it was not known as to why the telegraphy signal was distorted. Ideally, the signal should be a digital signal switching between one’s and zero’s, but the digital signal lost its shape rapidly along a telegraphy line.4
It was not until 1865 that James Clerk Maxwell [17] put in the missing term in Am- pere’s law, the displacement current term, only then the mathematical theory for electricity and magnetism was complete. Ampere’s law is now known as generalized Ampere’s law. The complete set of equations are now named Maxwell’s equations in honor of James Clerk Maxwell.
The rousing success of Maxwell’s theory was that it predicted wave phenomena, as they have been observed along telegraphy lines. But it was not until 23 years later that Heinrich Hertz in 1888 [18] did experiment to prove that electromagnetic field can propagate through space across a room. This illustrates the difficulty of knowledge dissemination when new knowledge is discovered. Moreover, from experimental measurement of the permittivity and permeability of matter, it was decided that electromagnetic wave moves at a tremendous speed. But the velocity of light has been known for a long while from astronomical obser- vations (Roemer, 1676) [19]. The observation of interference phenomena in light has been known as well. When these pieces of information were pieced together, it was decided that electricity and magnetism, and optics, are actually governed by the same physical law or Maxwell’s equations. And optics and electromagnetics are unified into one field!
4As a side note, in 1837, Morse invented the Morse code for telegraphy [16]. There were cross pollination of ideas across the Atlantic ocean despite the distance. In fact, Benjamin Franklin associated lightning with electricity in the latter part of the 18-th century. Also, notice that electrical machinery was invented in 1832 even though electromagnetic theory was not fully understood. 6 Electromagnetic Field Theory
Figure 1.3: A brief history of electromagnetics and optics as depicted in this figure.
In Figure 1.3, a brief history of electromagnetics and optics is depicted. In the begin- ning, it was thought that electricity and magnetism, and optics were governed by different physical laws. Low frequency electromagnetics was governed by the understanding of fields and their interaction with media. Optical phenomena were governed by ray optics, reflection and refraction of light. But the advent of Maxwell’s equations in 1865 revealed that they can be unified under electromagnetic theory. Then solving Maxwell’s equations becomes a mathematical endeavor. The photo-electric effect [20,21], and Planck radiation law [22] point to the fact that elec- tromagnetic energy is manifested in terms of packets of energy, indicating the corpuscular nature of light. Each unit of this energy is now known as the photon. A photon carries an en- ergy packet equal to ~ω, where ω is the angular frequency of the photon and ~ = 6.626×10−34 J s, the Planck constant, which is a very small constant. Hence, the higher the frequency, the easier it is to detect this packet of energy, or feel the graininess of electromagnetic en- ergy. Eventually, in 1927 [3], quantum theory was incorporated into electromagnetics, and the quantum nature of light gives rise to the field of quantum optics. Recently, even microwave photons have been measured [23,24]. They are difficult to detect because of the low frequency of microwave (109 Hz) compared to optics (1015 Hz): a microwave photon carries a packet of energy about a million times smaller than that of optical photon. The progress in nano-fabrication [25] allows one to make optical components that are subwavelength as the wavelength of blue light is about 450 nm. As a result, interaction of light with nano-scale optical components requires the solution of Maxwell’s equations in its full glory. Introduction, Maxwell’s Equations 7
In the early days of quantum theory, there were two prevailing theories of quantum in- terpretation. Quantum measurements were found to be random. In order to explain the probabilistic nature of quantum measurements, Einstein posited that a random hidden vari- able controlled the outcome of an experiment. On the other hand, the Copenhagen school of interpretation led by Niels Bohr, asserted that the outcome of a quantum measurement is not known until after a measurement [26]. In 1960s, Bell’s theorem (by John Steward Bell) [27] said that an inequality should be satisfied if Einstein’s hidden variable theory was correct. Otherwise, the Copenhagen school of interpretation should prevail. However, experimental measurement showed that the in- equality was violated, favoring the Copenhagen school of quantum interpretation [26]. This interpretation says that a quantum state is in a linear superposition of states before a mea- surement. But after a measurement, a quantum state collapses to the state that is measured. This implies that quantum information can be hidden incognito in a quantum state. Hence, a quantum particle, such as a photon, its state is unknown until after its measurement. In other words, quantum theory is “spooky”. This leads to growing interest in quantum infor- mation and quantum communication using photons. Quantum technology with the use of photons, an electromagnetic quantum particle, is a subject of growing interest. This also has the profound and beautiful implication that “our karma is not written on our forehead when we were born, our future is in our own hands!”
1.2 Maxwell’s Equations in Integral Form
Maxwell’s equations can be presented as fundamental postulates.5 We will present them in their integral forms, but will not belabor them until later. ¤ d E · dl = − B · dS Faraday’s Law (1.2.1) C dt¤ S d H · dl = D · dS + I Ampere’s Law (1.2.2) C dt S D · dS = Q Gauss’s or Coulomb’s Law (1.2.3) S B · dS = 0 Gauss’s Law (1.2.4) S The units of the basic quantities above are given as: E: V/m H: A/m D: C/m2 B: W/m2 I:A Q:C where V=volts, A=amperes, C=coulombs, and W=webers. 5Postulates in physics are similar to axioms in mathematics. They are assumptions that need not be proved. 8 Electromagnetic Field Theory 1.3 Static Electromagnetics 1.3.1 Coulomb’s Law (Statics)
This law, developed in 1785 [28], expresses the force between two charges q1 and q2. If these charges are positive, the force is repulsive and it is given by
q1q2 f = ˆr (1.3.1) 1→2 4πεr2 12
Figure 1.4: The force between two charges q1 and q2. The force is repulsive if the two charges have the same sign. where the units are: f (force): newton q (charge): coulombs ε (permittivity): farads/meter r (distance between q1 and q2): m ˆr12= unit vector pointing from charge 1 to charge 2
r2 − r1 ˆr12 = , r = |r2 − r1| (1.3.2) |r2 − r1| Since the unit vector can be defined in the above, the force between two charges can also be rewritten as
q1q2(r2 − r1) f1→2 = 3 , (r1, r2 are position vectors) (1.3.3) 4πε|r2 − r1|
1.3.2 Electric Field E (Statics) The electric field E is defined as the force per unit charge [29]. For two charges, one of charge q and the other one of incremental charge ∆q, the force between the two charges, according to Coulomb’s law (1.3.1), is q∆q f = ˆr (1.3.4) 4πεr2 Introduction, Maxwell’s Equations 9 where ˆr is a unit vector pointing from charge q to the incremental charge ∆q. Then the electric field E, which is the force per unit charge, is given by f E = , (V/m) (1.3.5) 4q
This electric field E from a point charge q at the orgin is hence q E = ˆr (1.3.6) 4πεr2 Therefore, in general, the electric field E(r) at location r from a point charge q at r0 is given by q(r − r0) E(r) = (1.3.7) 4πε|r − r0|3 where the unit vector r − r0 ˆr = (1.3.8) |r − r0|
Figure 1.5: Emanating E field from an electric point charge as depicted by depicted by (1.3.7) and (1.3.6).
If one knows E due to a point charge, one will know E due to any charge distribution because any charge distribution can be decomposed into sum of point charges. For instance, if 10 Electromagnetic Field Theory
there are N point charges each with amplitude qi, then by the principle of linear superposition, the total field produced by these N charges is
N X qi(r − ri) E(r) = (1.3.9) 4πε|r − r |3 i=1 i where qi = %(ri)∆Vi is the incremental charge at ri enclosed in the volume ∆Vi. In the continuum limit, one gets ¢ %(r0)(r − r0) E(r) = 0 3 dV (1.3.10) V 4πε|r − r | In other words, the total field, by the principle of linear superposition, is the integral sum- mation of the contributions from the distributed charge density %(r). Introduction, Maxwell’s Equations 11
1.3.3 Gauss’s Law (Statics)
This law is also known as Coulomb’s law as they are closely related to each other. Apparently, this simple law was first expressed by Joseph Louis Lagrange [30] and later, reexpressed by Gauss in 1813 (Wikipedia). This law can be expressed as D · dS = Q (1.3.11) S where D: electric flux density with unit C/m2 and D = εE. dS: an incremental surface at the point on S given by dSnˆ where nˆ is the unit normal pointing outward away from the surface. Q: total charge enclosed by the surface S.
Figure 1.6: Electric flux (courtesy of Ramo, Whinnery, and Van Duzer [31]).
The left-hand side of (1.3.11) represents a surface integral over a closed surface S. To understand it, one can break the surface into a sum of incremental surfaces ∆Si, with a local unit normal nˆi associated with it. The surface integral can then be approximated by a summation X X D · dS ≈ Di · nˆi∆Si = Di · ∆Si (1.3.12) S i i where one has defined ∆Si = nˆi∆Si. In the limit when ∆Si becomes infinitesimally small, the summation becomes a surface integral. 12 Electromagnetic Field Theory
1.3.4 Derivation of Gauss’s Law from Coulomb’s Law (Statics) From Coulomb’s law and the ensuing electric field due to a point charge, the electric flux is q D = εE = ˆr (1.3.13) 4πr2 When a closed spherical surface S is drawn around the point charge q, by symmetry, the electric flux though every point of the surface is the same. Moreover, the normal vector nˆ on the surface is just ˆr. Consequently, D · nˆ = D · ˆr = q/(4πr2), which is a constant on a spherical of radius r. Hence, we conclude that for a point charge q, and the pertinent electric flux D that it produces on a spherical surface, 2 2 D · dS = 4πr D · nˆ = 4πr Dr = q (1.3.14) S Therefore, Gauss’s law is satisfied by a point charge.
Figure 1.7: Electric flux from a point charge satisfies Gauss’s law.
Even when the shape of the spherical surface S changes from a sphere to an arbitrary shape surface S, it can be shown that the total flux through S is still q. In other words, the total flux through sufaces S1 and S2 in Figure 1.8 are the same. This can be appreciated by taking a sliver of the angular sector as shown in Figure 1.9. Here, ∆S1 and ∆S2 are two incremental surfaces intercepted by this sliver of angular sector. The amount of flux passing through this incremental surface is given by dS · D = nˆ · D∆S = nˆ · ˆrDr∆S. Here, D = ˆrDr is pointing in the ˆr direction. In ∆S1,n ˆ is pointing in the ˆr direction. But in ∆S2, the incremental area has been enlarged by that nˆ not aligned with D. But this enlargement is compensated by nˆ · ˆr. Also, ∆S2 has grown bigger, but the flux 2 at ∆S2 has grown weaker by the ratio of (r2/r1) . Finally, the two fluxes are equal in the limit that the sliver of angular sector becomes infinitesimally small. This proves the assertion that the total fluxes through S1 and S2 are equal. Since the total flux from a point charge q through a closed surface is independent of its shape, but always equal to q, then if we have a total charge Q which can be expressed as the sum of point charges, namely. X Q = qi (1.3.15) i Introduction, Maxwell’s Equations 13
Figure 1.8: Same amount of electric flux from a point charge passes through two surfaces S1 and S2.
Then the total flux through a closed surface equals the total charge enclosed by it, which is the statement of Gauss’s law or Coulomb’s law.
1.4 Homework Examples
Example 1
Field of a ring of charge of density %l C/m. Question: What is E along z axis? Hint: Use symmetry.
Example 2 Field between coaxial cylinders of unit length. Question: What is E? Hint: Use symmetry and cylindrical coordinates to express E =ρE ˆ ρ and appply Gauss’s law.
Example 3 Fields of a sphere of uniform charge density. Question: What is E? Hint: Again, use symmetry and spherical coordinates to express E =rE ˆ r and appply Gauss’s law. 14 Electromagnetic Field Theory
Figure 1.9: When a sliver of angular sector is taken, same amount of electric flux from a point charge passes through two incremental surfaces ∆S1 and ∆S2.
Figure 1.10: Electric field of a ring of charge (courtesy of Ramo, Whinnery, and Van Duzer [31]).
Figure 1.11: Figure for Example 2 for a coaxial cylinder. Introduction, Maxwell’s Equations 15
Figure 1.12: Figure for Example 3 for a sphere with uniform charge density. 16 Electromagnetic Field Theory Lecture 2
Maxwell’s Equations, Differential Operator Form
Maxwell’s equations were originally written in integral forms as has been shown in the previous lecture. Integral forms have nice physical meaning and can be easily related to experimental measurements. However, the differential operator forms1 can be easily converted to differential equations or partial differential equations where a whole sleuth of mathematical methods and numerical methods can be deployed. Therefore, it is prudent to derive the differential operator form of Maxwell’s equations.
2.1 Gauss’s Divergence Theorem
The divergence theorem is one of the most important theorems in vector calculus [31–34]. First, we will need to prove Gauss’s divergence theorem, namely, that: ¦ dV ∇ · D = D · dS (2.1.1) V S In the above, ∇ · D is defined as D · dS ∇ · D = lim ∆S (2.1.2) ∆V →0 ∆V The above implies that the divergence of the electric flux D, or ∇·D is given by first computing the flux coming (or oozing) out of a small volume ∆V surrounded by a small surface ∆S and taking their ratio as shown on the right-hand side. As shall be shown, the ratio has a limit and eventually, we will find an expression for it. We know that if ∆V ≈ 0 or small, then the
1We caution ourselves not to use the term “differential forms” which has a different meaning used in differential geometry for another form of Maxwell’s equations.
17 18 Electromagnetic Field Theory above,
∆V ∇ · D ≈ D · dS (2.1.3) ∆S
First, we assume that a volume V has been discretized2 into a sum of small cuboids, where the i-th cuboid has a volume of ∆Vi as shown in Figure 2.1. Then
N X V ≈ ∆Vi (2.1.4) i=1
Figure 2.1: The discretization of a volume V into a sum of small volumes ∆Vi each of which is a small cuboid. Stair-casing error occurs near the boundary of the volume V but the error diminishes as ∆Vi → 0.
2Other terms used are “tesselated”, “meshed”, or “gridded”. Maxwell’s Equations, Differential Operator Form 19
Figure 2.2: Fluxes from adjacent cuboids cancel each other leaving only the fluxes at the boundary that remain uncancelled. Please imagine that there is a third dimension of the cuboids in this picture where it comes out of the paper.
Then from (2.1.2),
∆Vi∇ · Di ≈ Di · dSi (2.1.5) ∆Si
By summing the above over all the cuboids, or over i, one gets X X ∆Vi∇ · Di ≈ Di · dSi ≈ D · dS (2.1.6) i i ∆Si S
It is easily seen that the fluxes out of the inner surfaces of the cuboids cancel each other, leaving only fluxes flowing out of the cuboids at the edge of the volume V as explained in Figure 2.2. The right-hand side of the above equation (2.1.6) becomes a surface integral over the surface S except for the stair-casing approximation (see Figure 2.1). However, this approximation becomes increasingly good as ∆Vi → 0. Moreover, the left-hand side becomes a volume integral, and we have ¦ dV∇ · D = D · dS (2.1.7) V S The above is Gauss’s divergence theorem.
2.1.1 Some Details Next, we will derive the details of the definition embodied in (2.1.2). To this end, we evaluate the numerator of the right-hand side carefully, in accordance to Figure 2.3. 20 Electromagnetic Field Theory
Figure 2.3: Figure to illustrate the calculation of fluxes from a small cuboid where a corner of the cuboid is located at (x0, y0, z0). There is a third z dimension of the cuboid not shown, and coming out of the paper. Hence, this cuboid, unlike that shown in the figure, has six faces.
Accounting for the fluxes going through all the six faces, assigning the appropriate signs in accordance with the fluxes leaving and entering the cuboid, one arrives at
D · dS ≈ −Dx(x0, y0, z0)∆y∆z + Dx(x0 + ∆x, y0, z0)∆y∆z ∆S −Dy(x0, y0, z0)∆x∆z + Dy(x0, y0 + ∆y, z0)∆x∆z
−Dz(x0, y0, z0)∆x∆y + Dz(x0, y0, z0 + ∆z)∆x∆y (2.1.8)
Factoring out the volume of the cuboid ∆V = ∆x∆y∆z in the above, one gets
D · dS ≈ ∆V {[Dx(x0 + ∆x, . . .) − Dx(x0,...)] /∆x ∆S + [Dy(. . . , y0 + ∆y, . . .) − Dy(. . . , y0,...)] /∆y
+ [Dz(. . . , z0 + ∆z) − Dz(. . . , z0)] /∆z} (2.1.9)
Or that
D · dS ∂D ∂D ∂D ≈ x + y + z (2.1.10) ∆V ∂x ∂y ∂z In the limit when ∆V → 0, then
D · dS ∂D ∂D ∂D lim = x + y + z = ∇ · D (2.1.11) ∆V →0 ∆V ∂x ∂y ∂z Maxwell’s Equations, Differential Operator Form 21 where ∂ ∂ ∂ ∇ =x ˆ +y ˆ +z ˆ (2.1.12) ∂x ∂y ∂z
D =xD ˆ x +yD ˆ y +zD ˆ z (2.1.13) The above is the definition of the divergence operator in Cartesian coordinates. The diver- gence operator ∇· has its complicated representations in cylindrical and spherical coordinates, a subject that we would not delve into in this course. But they can be derived, and are best looked up at the back of some textbooks on electromagnetics. Consequently, one gets Gauss’s divergence theorem given by ¦ dV ∇ · D = D · dS (2.1.14) V S
2.1.2 Gauss’s Law in Differential Operator Form By further using Gauss’s or Coulomb’s law implies that ¦ D · dS = Q = dV % (2.1.15) S We can esplace the left-hand side of the above by (2.1.14) to arrive at ¦ ¦ dV ∇ · D = dV % (2.1.16) V V When V → 0, we arrive at the pointwise relationship, a relationship at an arbitrary point in space. Therefore,
∇ · D = % (2.1.17)
2.1.3 Physical Meaning of Divergence Operator The physical meaning of divergence is that if ∇·D 6= 0 at a point in space, it implies that there are fluxes oozing or exuding from that point in space [35]. On the other hand, if ∇ · D = 0, it implies no flux oozing out from that point in space. In other words, whatever flux that goes into the point must come out of it. The flux is termed divergence free. Thus, ∇ · D is a measure of how much sources or sinks exist for the flux at a point. The sum of these sources or sinks gives the amount of flux leaving or entering the surface that surrounds the sources or sinks. Moreover, if one were to integrate a divergence-free flux over a volume V , and invoking Gauss’s divergence theorem, one gets D · dS = 0 (2.1.18) S In such a scenerio, whatever flux that enters the surface S must leave it. In other words, what comes in must go out of the volume V , or that flux is conserved. This is true of incompressible 22 Electromagnetic Field Theory
fluid flow, electric flux flow in a source free region, as well as magnetic flux flow, where the flux is conserved.
Figure 2.4: In an incompressible flux flow, flux is conserved: whatever flux that enters a volume V must leave the volume V .
2.2 Stokes’s Theorem
The mathematical description of fluid flow was well established before the establishment of electromagnetic theory [36]. Hence, much mathematical description of electromagnetic theory uses the language of fluid. In mathematical notations, Stokes’s theorem is ¤ E · dl = ∇ × E · dS (2.2.1) C S
In the above, the contour C is a closed contour, whereas the surface S is not closed.3 First, applying Stokes’s theorem to a small surface ∆S, we define a curl operator4 ∇× at a point to be measured as
E · dl (∇ × E) · nˆ = lim ∆C (2.2.2) ∆S→0 ∆S
In the above, ∇ × E is a vector. Taking ∆C E · dl as a measure of the rotation of the field E around a small loop ∆C, the ratio of this rotation to the area of the loop ∆S has a limit when ∆S becomes infinitesimally small. This ratio is related to ∇ × E.
3In other words, C has no boundary whereas S has boundary. A closed surface S has no boundary like when we were proving Gauss’s divergence theorem previously. 4Sometimes called a rotation operator. Maxwell’s Equations, Differential Operator Form 23
Figure 2.5: In proving Stokes’s theorem, a closed contour C is assumed to enclose an open surface S. Then the surface S is tessellated into sum of small rects as shown. Stair-casing error vanishes in the limit when the rects are made vanishingly small.
First, the surface S enclosed by C is tessellated (also called meshed, gridded, or discretized) into sum of small rects (rectangles) as shown in Figure 2.5. Stokes’s theorem is then applied to one of these small rects to arrive at
Ei · dli = (∇ × Ei) · ∆Si (2.2.3) ∆Ci
where one defines ∆Si =n ˆ∆S. Next, we sum the above equation over i or over all the small rects to arrive at
X X Ei · dli = ∇ × Ei · ∆Si (2.2.4) i ∆Ci i
Again, on the left-hand side of the above, all the contour integrals over the small rects cancel each other internal to S save for those on the boundary. In the limit when ∆Si → 0, the left-hand side becomes a contour integral over the larger contour C, and the right-hand side becomes a surface integral over S. One arrives at Stokes’s theorem, which is
¤ E · dl = (∇ × E) · dS (2.2.5) C S 24 Electromagnetic Field Theory
Figure 2.6: We approximate the integration over a small rect using this figure. There are four edges to this small rect.
Next, we need to prove the details of definition (2.2.2) using Figure 2.6. Performing the integral over the small rect, one gets
E · dl = Ex(x0, y0, z0)∆x + Ey(x0 + ∆x, y0, z0)∆y ∆C − Ex(x0, y0 + ∆y, z0)∆x − Ey(x0, y0, z0)∆y E (x , y , z ) E (x , y + ∆y, z ) = ∆x∆y x 0 0 0 − x 0 0 0 ∆y ∆y E (x , y , z ) E (x , y + ∆y, z ) − y 0 0 0 + y 0 0 0 ∆x ∆x (2.2.6)
We have picked the normal to the incremental surface ∆S to bez ˆ in the above example, and hence, the above gives rise to the identity that
∆S E · dl ∂ ∂ lim = Ey − Ex =z ˆ · ∇ × E (2.2.7) ∆S→0 ∆S ∂x ∂y Picking different ∆S with different orientations and normalsn ˆ where |hatn =x ˆ orn ˆ =y ˆ, one gets ∂ ∂ E − E =x ˆ · ∇ × E (2.2.8) ∂y z ∂z y ∂ ∂ E − E =y ˆ · ∇ × E (2.2.9) ∂z x ∂x z Maxwell’s Equations, Differential Operator Form 25
The above gives the x, y, and z components of ∇ × E. It is to be noted that ∇ × E is a vector. In other words, one gets
∂ ∂ ∂ ∂ ∇ × E =x ˆ E − E +y ˆ E − E ∂y z ∂z y ∂z x ∂x z ∂ ∂ +ˆz E − E (2.2.10) ∂x y ∂y x where ∂ ∂ ∂ ∇ =x ˆ +y ˆ +z ˆ (2.2.11) ∂x ∂y ∂z
2.2.1 Faraday’s Law in Differential Operator Form Faraday’s law is experimentally motivated. Michael Faraday (1791-1867) was an extraordi- nary experimentalist who documented this law with meticulous care. It was only decades later that a mathematical description of this law was arrived at. Faraday’s law in integral form is given by5 ¤ d E · dl = − B · dS (2.2.12) C dt S Assuming that the surface S is not time varying, one can take the time derivative into the integrand and write the above as ¤ ∂ E · dl = − B · dS (2.2.13) C S ∂t One can replace the left-hand side with the use of Stokes’ theorem to arrive at ¤ ¤ ∂ ∇ × E · dS = − B · dS (2.2.14) S S ∂t The normal of the surface element dS can be pointing in an arbitrary direction, and the surface S can be made very small. Then the integral can be removed, and one has
∂ ∇ × E = − B (2.2.15) ∂t The above is Faraday’s law in differential operator form. ∂B In the static limit, ∂t = 0, giving
∇ × E = 0 (2.2.16)
5Faraday’s law is experimentally motivated. Michael Faraday (1791-1867) was an extraordinary exper- imentalist who documented this law with meticulous care. It was only decades later that a mathematical description of this law was arrived at. 26 Electromagnetic Field Theory
2.2.2 Physical Meaning of Curl Operator
The curl operator ∇× is a measure of the rotation or the circulation of a field at a point in space. On the other hand, ∆C E · dl is a measure of the circulation of the field E around the loop formed by C. Again, the curl operator has its complicated representations in other co- ordinate systems like cylindrical or spherical coordinates, a subject that will not be discussed in detail here. It is to be noted that our proof of the Stokes’s theorem is for a flat open surface S, and not for a general curved open surface. Since all curved surfaces can be tessellated into a union of flat triangular surfaces according to the tiling theorem, the generalization of the above proof to curved surface is straightforward. An example of such a triangulation of a curved surface into a union of flat triangular surfaces is shown in Figure 2.7.
Figure 2.7: An arbitrary curved surface can be triangulated with flat triangular patches. The triangulation can be made arbitrarily accurate by making the patches arbitrarily small.
2.3 Maxwell’s Equations in Differential Operator Form
With the use of Gauss’ divergence theorem and Stokes’ theorem, Maxwell’s equations can be written more elegantly in differential operator forms. They are:
∂B ∇ × E = − (2.3.1) ∂t ∂D ∇ × H = + J (2.3.2) ∂t ∇ · D = % (2.3.3) ∇ · B = 0 (2.3.4)
These equations are point-wise relations as they relate the left-hand side and right-hand side field values at a given point in space. Moreover, they are not independent of each other. For instance, one can take the divergence of the first equation (2.3.1), making use of the vector Maxwell’s Equations, Differential Operator Form 27 identity that ∇ · (∇ × E) = 0, one gets
∂∇ · B − = 0 → ∇ · B = constant (2.3.5) ∂t This constant corresponds to magnetic charges, and since they have not been experimentally observed, one can set the constant to zero. Thus the fourth of Maxwell’s equations, (2.3.4), follows from the first (2.3.1). Similarly, by taking the divergence of the second equation (2.3.2), and making use of the current continuity equation that
∂% ∇ · J + = 0 (2.3.6) ∂t one can obtain the second last equation (2.3.3). Notice that in (2.3.3), the charge density % can be time-varying, whereas in the previous lecture, we have “derived” this equation from Coulomb’s law using electrostatic theory. The above logic follows if ∂/∂t 6= 0, and is not valid for static case. In other words, for statics, the third and the fourth equations are not derivable from the first two. Hence all four Maxwell’s equations are needed for static problems. For electrodynamic problems, only solving the first two suffices. Something is amiss in the above. If J is known, then solving the first two equations implies solving for four vector unknowns, E, H, B, D, which has 12 scalar unknowns. But there are only two vector equations or 6 scalar equations in the first two equations. Thus one needs more equations. These are provided by the constitutive relations that we shall discuss next.
2.4 Homework Examples
Example 1 If D = (2y2 + z)ˆx + 4xyyˆ + xzˆ, find:
1. Volume charge density ρv at (−1, 0, 3).
2. Electric flux through the cube defined by
0 ≤ x ≤ 1, 0 ≤ y ≤ 1, 0 ≤ z ≤ 1.
3. Total charge enclosed by the cube.
Example 2 ¢ Suppose E = ˆx3y + ˆyx, calculate E · dl along a straight line in the x-y plane joining (0,0) to (3,1). 28 Electromagnetic Field Theory 2.5 Historical Notes
There are several interesting historical notes about Maxwell.
It is to be noted that when James Clerk Maxwell first wrote his equations down, it was in many equations and very difficult to digest [17, 37, 38]. It was Oliver Heaviside who distilled those equations and presented them in the present form found in textbooks. Putatively, most cannot read his treatise [37] beyond the first 50 pages [39]. Maxwell wrote many poems in his short lifespan (1831-1879) and they can be found at [40]. Also, the ancestor of James Clerk Maxwell married from the Clerk family into the Maxwell family. One of the conditions of marriage was that all the descendants of the Clerk family would be named Clerk Maxwell. That was why Maxwell is addressed as Professor Clerk Maxwell. Lecture 3
Constitutive Relations, Wave Equation, Electrostatics, and Static Green’s Function
Constitutive relations are important for defining the electromagnetic material properties of the media involved. Also, wave phenomenon is a major triumph of Maxwell’s equations. Hence, we will study that derivation of this phenomenon here. To make matter simple, we will reduce the problem to electrostatics to simplify the math to introduce the concept of the Green’s function. As mentioned previously, for time-varying problems, only the first two of the four Maxwell’s equations suffice. The latter two are derivable from the first two. But the first two equations have four unknowns E, H, D, and B. Hence, two more equations are needed to solve for these unknowns. These equations come from the constitutive relations.
3.1 Simple Constitutive Relations
The constitution relation between electric flux D and the electric field E in free space (or vacuum) is
D = ε0E (3.1.1)
When material medium is present, one has to add the contribution to D by the polarization density P which is a dipole density.1 Then [31, 32,41]
D = ε0E + P (3.1.2)
1Note that a dipole moment is given by Q` where Q is its charge in coulomb and ` is its length in m. Hence, dipole density, or polarization density as dimension of coulomb/m2, which is the same as that of electric flux D.
29 30 Electromagnetic Field Theory
The second term P above is due to material property, and the contribution to the electric flux due to the polarization density of the medium. It is due to the little dipole contribution due to the polar nature of the atoms or molecules that make up a medium. By the same token, the first term ε0E can be thought of as the polarization density contribution of vacuum. Vacuum, though represents nothingness, has electrons and positrons, or electron-positron pairs lurking in it [42]. Electron is matter, whereas positron is anti- matter. In the quiescent state, they represent nothingness, but they can be polarized by an electric field E. That also explains why electromagnetic wave can propagate through vacuum. For many media, they are approximately linear media. Then P is linearly proportional to E, or P = ε0χ0E, or
D = ε0E + ε0χ0E
= ε0(1 + χ0)E = εE, ε = ε0(1 + χ0) = ε0εr (3.1.3) where χ0 is the electric susceptibility . In other words, for linear material media, one can replace the vacuum permittivity ε0 with an effective permittivity ε = ε0εr where εr is the relative permittivity. Thus, D is linearly proportional to E. In free space,2
10−8 ε = ε = 8.854 × 10−12 ≈ F/m (3.1.4) 0 36π The constitutive relation between magnetic flux B and magnetic field H is given as
B = µH, µ = permeability H/m (3.1.5)
In free space or vacuum,
−7 µ = µ0 = 4π × 10 H/m (3.1.6)
As shall be explained later, this is an assigned value giving it a precise value as shown above. In other materials, the permeability can be written as
µ = µ0µr (3.1.7)
The above can be derived using similar argument for permittivity, where the different perme- ability is due to the presence of magnetic dipole density in a material medium. In the above, µr is termed the relative permeability.
3.2 Emergence of Wave Phenomenon, Triumph of Maxwell’s Equations
One of the major triumphs of Maxwell’s equations is the prediction of the wave phenomenon. This was experimentally verified by Heinrich Hertz in 1888 [18], some 23 years after the
2It is to be noted that we will use MKS unit in this course. Another possible unit is the CGS unit used in many physics texts [43] Constitutive Relations, Wave Equation, Electrostatics, and Static Green’s Function 31 completion of Maxwell’s theory in 1865 [17]. To see this, we consider the first two Maxwell’s equations for time-varying fields in vacuum or a source-free medium.3 They are ∂H ∇ × E = −µ (3.2.1) 0 ∂t ∂E ∇ × H = −ε (3.2.2) 0 ∂t Taking the curl of (3.2.1), we have
∂ ∇ × ∇ × E = −µ ∇ × H (3.2.3) 0 ∂t It is understood that in the above, the double curl operator implies ∇×(∇×E). Substituting (3.2.2) into (3.2.3), we have
∂2 ∇ × ∇ × E = −µ ε E (3.2.4) 0 0 ∂t2 In the above, the left-hand side can be simplified by using the identity that a × (b × c) = b(a · c) − c(a · b),4 but be mindful that the operator ∇ has to operate on a function to its right. Therefore, we arrive at the identity that
∇ × ∇ × E = ∇∇ · E − ∇2E (3.2.5)
Since ∇ · E = 0 in a source-free medium, we have
∂2 ∇2E − µ ε E = 0 (3.2.6) 0 0 ∂t2 where ∂2 ∂2 ∂2 ∇2 = ∇ · ∇ = + + ∂x2 ∂y2 ∂z2 The above is known as the Laplacian operator. Here, (3.2.6) is the wave equation in three space dimensions [32,44]. To see the simplest form of wave emerging in the above, we can let E =xE ˆ x(z, t) so that ∇ · E = 0 satisfying the source-free condition. Then (3.2.6) becomes
∂2 ∂2 E (z, t) − µ ε E (z, t) = 0 (3.2.7) ∂z2 x 0 0 ∂t2 x Eq. (3.2.7) is known mathematically as the wave equation in one space dimension. It can also be written as ∂2 1 ∂2 2 f(z, t) − 2 2 f(z, t) = 0 (3.2.8) ∂z c0 ∂t
3 Since the third and the fourth Maxwell’s equations are derivable from the first two when ∂/∂t 6= 0. 4For mnemonics, this formula is also known as the “back-of-the-cab” formula. 32 Electromagnetic Field Theory
2 −1 where c0 = (µ0ε0) . Eq. (3.2.8) can also be factorized as
∂ 1 ∂ ∂ 1 ∂ − + f(z, t) = 0 (3.2.9) ∂z c0 ∂t ∂z c0 ∂t or
∂ 1 ∂ ∂ 1 ∂ + − f(z, t) = 0 (3.2.10) ∂z c0 ∂t ∂z c0 ∂t
The above can be verified easily by direct expansion, and using the fact that
∂ ∂ ∂ ∂ = (3.2.11) ∂t ∂z ∂z ∂t
The above implies that we have either
∂ 1 ∂ + f+(z, t) = 0 (3.2.12) ∂z c0 ∂t or
∂ 1 ∂ − f−(z, t) = 0 (3.2.13) ∂z c0 ∂t
Equation (3.2.12) and (3.2.13) are known as the one-way wave equations or the advective equations [45]. From the above factorization, it is seen that the solutions of these one-way wave equations are also the solutions of the original wave equation given by (3.2.8). Their general solutions are then
f+(z, t) = F+(z − c0t) (3.2.14)
f−(z, t) = F−(z + c0t) (3.2.15)
We can verify the above by back substitution into (3.2.12) and (3.2.13). Eq. (3.2.14) consti- tutes a right-traveling wave function of any shape while (3.2.15) constitutes a left-traveling wave function of any shape. Since Eqs. (3.2.14) and (3.2.15) are also solutions to (3.2.8), we can write the general solution to the wave equation as
f(z, t) = F+(z − c0t) + F−(z + c0t) (3.2.16)
This is a wonderful result since F+ and F− are arbitrary functions of any shape (see Figure 3.1); they can be used to encode information for communication! Constitutive Relations, Wave Equation, Electrostatics, and Static Green’s Function 33
Figure 3.1: Solutions of the wave equation can be a single-valued function of any shape. In the above, F+ travels in the positive z direction, while F− travels in the negative z direction as t increases.
Furthermore, one can calculate the velocity of this wave to be
8 c0 = 299, 792, 458m/s ' 3 × 10 m/s (3.2.17) p where c0 = 1/µ0ε0. Maxwell’s equations (3.2.1) implies that E and H are linearly proportional to each other. Thus, there is only one independent constant in the wave equation, and the value of µ0 is −7 −1 defined neatly to be 4π × 10 henry m , while the value of ε0 has been measured to be about 8.854 × 10−12 farad m−1. Now it has been decided that the velocity of light is used as a standard and is defined to be the integer given in (3.2.17). A meter is defined to be the distance traveled by light in 1/(299792458) seconds. Hence, the more accurate that unit of time or second can be calibrated, the more accurate can we calibrate the unit of length or meter. Therefore, the design of an accurate clock like an atomic clock is an important research problem. The value of ε0 was measured in the laboratory quite early. Then it was realized that electromagnetic wave propagates at a tremendous velocity which is the velocity of light.5 This
5The velocity of light was known in astronomy by (Roemer, 1676) [19]. 34 Electromagnetic Field Theory was also the defining moment which revealed that the field of electricity and magnetism and the field of optics were both described by Maxwell’s equations or electromagnetic theory.
3.3 Static Electromagnetics–Revisted
We have seen static electromagnetics previously in integral form. Now we look at them in differential operator form. When the fields and sources are not time varying, namely that ∂/∂t = 0, we arrive at the static Maxwell’s equations for electrostatics and magnetostatics, namely [31, 32,46]
∇ × E = 0 (3.3.1) ∇ × H = J (3.3.2) ∇ · D = % (3.3.3) ∇ · B = 0 (3.3.4)
Notice the the electrostatic field system is decoupled from the magnetostatic field system. However, in a resistive system where
J = σE (3.3.5) the two systems are coupled again. This is known as resistive coupling between them. But if σ → ∞, in the case of a perfect conductor, or superconductor, then for a finite J, E has to be zero. The two systems are decoupled again. Also, one can arrive at the equations above by letting µ0 → 0 and 0 → 0. In this case, the velocity of light becomes infinite, or retardation effect is negligible. In other words, there is no time delay for signal propagation through the system in the static approximation. Finally, it is important to note that in statics, the latter two Maxwell’s equations are not derivable from the first two. Hence, all four equations have to be considered when one seeks the solution in the static regime or the long-wavelength regime.
3.3.1 Electrostatics We see that Faraday’s law in the static limit is
∇ × E = 0 (3.3.6)
One way to satisfy the above is to let E = −∇Φ because of the identity ∇ × ∇ = 0.6 Alternatively, one can assume that E is a constant. But we usually are interested in solutions that vanish at infinity, and hence, the latter is not a viable solution. Therefore, we let
E = −∇Φ (3.3.7)
6One an easily go through the algebra in cartesian coordinates to convince oneself of this. Constitutive Relations, Wave Equation, Electrostatics, and Static Green’s Function 35
3.3.2 Poisson’s Equation As a consequence of the above,
∇ · D = % ⇒ ∇ · εE = % ⇒ −∇ · ε∇Φ = % (3.3.8)
In the last equation above, if ε is a constant of space, or independent of r, then one arrives at the simple Poisson’s equation, which is a partial differential equation % ∇2Φ = − (3.3.9) ε Here, the Laplacian operator
∂2 ∂2 ∂2 ∇2 = ∇ · ∇ = + + ∂x2 ∂y2 ∂z2
For a point source, we know from Coulomb’s law that q E = rˆ = −∇Φ (3.3.10) 4πεr2 From the above, we deduce that7 q Φ = (3.3.11) 4πεr Therefore, we know the solution to Poisson’s equation (3.3.9) when the source % represents a point source. Since this is a linear equation, we can use the principle of linear superposition to find the solution when the charge density %(r) is arbitrary. A point source located at r0 is described by a charge density as
%(r) = qδ(r − r0) (3.3.12) where δ(r−r0) is a short-hand notation for δ(x−x0)δ(y −y0)δ(z −z0). Therefore, from (3.3.9), the corresponding partial differential equation for a point source is
qδ(r − r0) ∇2Φ(r) = − (3.3.13) ε The solution to the above equation, from Coulomb’s law in accordance to (3.3.11), has to be q Φ(r) = (3.3.14) 4πε|r − r0| whereas (3.3.11) is for a point source at the origin, (3.3.14) is for a point source located and translated to r0. The above is a coordinate independent form of the solution. Here, r =xx ˆ +yy ˆ +zz ˆ and r0 =xx ˆ 0 +yy ˆ 0 +zz ˆ 0, and |r − r0| = p(x − x0)2 + (y − y0)2 + (z − z0)2.
7One can always take the gradient or ∇ of Φ to verify this. Mind you, this is best done in spherical coordinates. 36 Electromagnetic Field Theory
3.3.3 Static Green’s Function Let us define a partial differential equation given by ∇2g(r − r0) = −δ(r − r0) (3.3.15) The above is similar to Poisson’s equation with a point source on the right-hand side as in (3.3.13). Such a solution, a response to a point source, is called the Green’s function.8 By comparing equations (3.3.13) and (3.3.15), then making use of (3.3.14), we deduced that the static Green’s function is 1 g(r − r0) = (3.3.16) 4π|r − r0| An arbitrary source can be expressed as ¦ %(r) = dV 0%(r0)δ(r − r0) (3.3.17) V The above is just the statement that an arbitrary charge distribution %(r) can be expressed as a linear superposition of point sources δ(r − r0). Using the above in (3.3.9), we have ¦ 1 ∇2Φ(r) = − dV 0%(r0)δ(r − r0) (3.3.18) ε V We can let ¦ 1 Φ(r) = dV 0g(r − r0)%(r0) (3.3.19) ε V By substituting the above into the left-hand side of (3.3.18), exchanging order of integration and differentiation, and then making use of equation (3.3.15), it can be shown that (3.3.19) indeed satisfies (3.3.9). The above is just a convolutional integral. Hence, the potential Φ(r) due to an arbitrary source distribution %(r) can be found by using convolution, namely, ¦ 0 1 %(r ) 0 Φ(r) = 0 dV (3.3.20) 4πε V |r − r | In a nutshell, the solution of Poisson’s equation when it is driven by an arbitrary source %, is the convolution of the source with the static Green’s function, a point source response.
3.3.4 Laplace’s Equation If % = 0, or if we are in a source-free region, then ∇2Φ =0 (3.3.21) which is the Laplace’s equation. Laplace’s equation is usually solved as a boundary value problem. In such a problem, the potential Φ is stipulated on the boundary of a region with a certain boundary condition, and then the solution is sought in the region so as to match the boundary condition. Examples of such boundary value problems are given at the end of the lecture.
8George Green (1793-1841), the son of a Nottingham miller, was self-taught, but his work has a profound impact in our world. Constitutive Relations, Wave Equation, Electrostatics, and Static Green’s Function 37
3.4 Homework Examples
Example 1
Fields of a sphere of radius a with uniform charge density ρ:
Figure 3.2: Figure of a sphere with uniform charge density for the example above.
Assuming that Φ|r=∞ = 0, what is Φ at r ≤ a? And Φ at r > a.
Example 2
A capacitor has two parallel plates attached to a battery, what is E field inside the capacitor? First, one guess the electric field between the two parallel plates. Then one arrive at a potential Φ in between the plates so as to produce the field. Then the potential is found so as to match the boundary conditions of Φ = V in the upper plate, and Φ = 0 in the lower plate. What is the Φ that will satisfy the requisite boundary condition?
Figure 3.3: Figure of a parallel plate capacitor. The field in between can be found by solving Laplace’s equation as a boundary value problem [31].
Example 3
A coaxial cable has two conductors. The outer conductor is grounded and hence is at zero potential. The inner conductor is at voltage V . What is the solution? 38 Electromagnetic Field Theory
For this, one will have to write the Laplace’s equation in cylindrical coordinates, namely,
1 ∂ ∂Φ 1 ∂2Φ ∇2Φ = ρ + = 0 (3.4.1) ρ ∂ρ ∂ρ ρ2 ∂φ2
In the above, we assume that the potential is constant in the z direction, and hence, ∂/∂z = 0, and ρ, φ, z are the cylindrical coordinates. By assuming axi-symmetry, we can let ∂/∂φ = 0 and the above becomes 1 ∂ ∂Φ ∇2Φ = ρ = 0 (3.4.2) ρ ∂ρ ∂ρ
Show that Φ = A ln ρ + B is a general solution to Laplace’s equation in cylindrical coor- dinates inside a coax. What is the Φ that will satisfy the requisite boundary condition?
Figure 3.4: The field in between a coaxial line can also be obtained by solving Laplace’s equation as a boundary value problem (courtesy of Ramo, Whinnery, and Van Duzer [31]). Lecture 4
Magnetostatics, Boundary Conditions, and Jump Conditions
In the previous lecture, Maxwell’s equations become greatly simplified in the static limit. We have looked at how the electrostatic problems are solved. We now look at the magnetostatic case. In addition, we will study boundary conditions and jump conditions at an interface, and how they are derived from Maxwell’s equations. Maxwell’s equations can be first solved in different domains. Then the solutions are pieced (or sewn) together by imposing boundary conditions at the boundaries or interfaces of the domains. Such problems are called boundary- value problems (BVPs).
4.1 Magnetostatics
From Maxwell’s equations, we can deduce that the magnetostatic equations for the magnetic field and flux when ∂/∂t = 0, which are [31, 32,46]
∇ × H = J (4.1.1) ∇ · B = 0 (4.1.2)
These two equations are greatly simplified, and hence, are easier to solve compared to the time-varying case. One way to satisfy the second equation is to let
B = ∇ × A (4.1.3) because of the vector identity
∇ · (∇ × A) = 0 (4.1.4)
39 40 Electromagnetic Field Theory
The above is zero for the same reason that a · (a × b) = 0. In this manner, Gauss’s law in (4.1.2) is automatically satisfied. From (4.1.1), we have
B ∇ × = J (4.1.5) µ
Then using (4.1.3) into the above,
1 ∇ × ∇ × A = J (4.1.6) µ
In a homogeneous medium,1 µ or 1/µ is a constant and it commutes with the differential ∇ operator or that it can be taken outside the differential operator. As such, one arrives at
∇ × (∇ × A) = µJ (4.1.7)
We use the vector identity that (see back-of-cab formula in the previous lecture)
∇ × (∇ × A) = ∇(∇ · A) − (∇ · ∇)A = ∇(∇ · A) − ∇2A (4.1.8) where ∇2 is a shorthand notation for ∇ · ∇. As a result, we arrive at [47]
∇(∇ · A) − ∇2A = µJ (4.1.9)
By imposing the Coulomb gauge that ∇ · A = 0, which will be elaborated in the next section, we arrive at the simplified equation
∇2A = −µJ (4.1.10)
The above is also known as the vector Poisson’s equation. In cartesian coordinates, the above can be viewed as three scalar Poisson’s equations. Each of the Poisson’s equation can be solved using the Green’s function method previously described. Consequently, in free space ¦ µ J(r0) A(r) = dV 0 (4.1.11) 4π V R where
R = |r − r0| (4.1.12) is the distance between the source point r0 and the observation point r. Here, dV 0 = dx0dy0dz0. It is also variously written as dr0 or d3r0.
1Its prudent to warn the reader of the use of the word “homogeneous”. In the math community, it usually refers to something to be set to zero. But in the electromagnetics community, it refers to something “non- heterogeneous”. Magnetostatics, Boundary Conditions, and Jump Conditions 41
4.1.1 More on Coulomb Gauge Gauge is a very important concept in physics [48], and we will further elaborate it here. First, notice that A in (4.1.3) is not unique because one can always define
A0 = A − ∇Ψ (4.1.13)
Then
∇ × A0 = ∇ × (A − ∇Ψ) = ∇ × A = B (4.1.14) where we have made use of that ∇ × ∇Ψ = 0. Hence, the ∇× of both A and A0 produce the same B; hence, A is non-unique. To find A uniquely, we have to define or set the divergence of A or provide a gauge condition. One way is to set the divergence of A to be zero, namely that
∇ · A = 0 (4.1.15)
Then
∇ · A0 = ∇ · A − ∇2Ψ 6= ∇ · A (4.1.16)
The last non-equal sign follows if ∇2Ψ 6= 0. However, if we further stipulate that ∇ · A0 = ∇ · A = 0, then −∇2Ψ = 0. This does not necessary imply that Ψ = 0, but if we impose that condition that Ψ → 0 when r → ∞, then Ψ = 0 everywhere.2 By so doing, A and A0 are equal to each other, and we obtain (4.1.10) and (4.1.11). The above is akin to the idea that given a vector a, just by stipulating that b × a = c is not enough to determine a. We need to stipulate what b · a is as well. Here, a, b, and c are independent vectors.
4.2 Boundary Conditions–1D Poisson’s Equation
To simplify the solutions of Maxwell’s equations, they are usually solved in a homogeneous medium. As mentioned before, a complex problem can be divided into piecewise homogeneous regions first, and then the solution in each region sought. Then the total solution must satisfy boundary conditions at the interface between the piecewise homogeneous regions. What are these boundary conditions? Boundary conditions are actually embedded in the partial differential equations that the potential or the field satisfy. Two important concepts to keep in mind are:
Differentiation of a function with discontinuous slope will give rise to step discontinuity. Differentiation of a function with step discontinuity will give rise to a Dirac delta func- tion. This is also called the jump condition, a term often used by the mathematics community [49].
2It is a property of the Laplace boundary value problem that if Ψ = 0 on a closed surface S, then Ψ = 0 everywhere inside S. Earnshaw’s theorem [31] is useful for proving this assertion. 42 Electromagnetic Field Theory
Take for example a one dimensional Poisson’s equation that d d ε(x) Φ(x) = −%(x) (4.2.1) dx dx where ε(x) represents material property that has the form given in Figure 4.1. One can actually say something about Φ(x) given %(x) on the right-hand side. If %(x) has a delta d function singularity, it implies that ε(x) dx Φ(x) has a step discontinuity. If %(x) is finite d everywhere, then ε(x) dx Φ(x) must be continuous everywhere. d Furthermore, if ε(x) dx Φ(x) is finite everywhere, it implies that Φ(x) must be continuous everywhere.
Figure 4.1: A figure showing a charge sheet at the interface between two dielectric media. Because it is a surface charge sheet, the volume charge density %(x) is infinite at the sheet location x0.
To see this in greater detail, we illustrate it with the following example. In the above, %(x) represents a charge distribution given by %(x) = %sδ(x − x0). In this case, the charge distribution is everywhere zero except at the location of the surface charge sheet, where the charge density is infinite: it is represented mathematically by a delta function3 in space. To find the boundary condition of the potential Φ(x) at x0 , we integrate (4.2.1) over an infinitesimal width around x0, the location of the charge sheet, namely ¢ ¢ ¢ x0+∆ d d x0+∆ x0+∆ dx ε(x) Φ(x) = − dx%(x) = − dx%0δ(x − x0) (4.2.2) x0−∆ dx dx x0−∆ x0−∆ Since the integrand of the left-hand side is an exact derivative, we get
x0+∆ d ε(x) Φ(x) = −%s (4.2.3) dx x0−∆ 3This function has been attributed to Dirac who used in pervasively, but Cauchy was aware of such a function. Magnetostatics, Boundary Conditions, and Jump Conditions 43 whereas on the right-hand side, we pick up the contribution from the delta function. Evalu- ating the left-hand side at their limits, one arrives at
d d ε(x+) Φ(x+) − ε(x−) Φ(x−) =∼ −% , (4.2.4) 0 dx 0 0 dx 0 s
± d where x0 = lim∆→0 x0 ± ∆. In other words, the jump discontinuity is in ε(x) dx Φ(x) and the amplitude of the jump discontinuity is proportional to the amplitude of the delta function. Since E = −∇Φ, or that
d E (x) = − Φ(x), (4.2.5) x dx
The above implies that
+ + − − ε(x0 )Ex(x0 ) − ε(x0 )Ex(x0 ) = %s (4.2.6) or
+ − Dx(x0 ) − Dx(x0 ) = %s (4.2.7) where
Dx(x) = ε(x)Ex(x) (4.2.8)
The lesson learned from above is that boundary condition is obtained by integrating the pertinent differential equation over an infinitesimal small segment. In this mathematical way of looking at the boundary condition, one can also eyeball the differential equation and ascertain the terms that will have the jump discontinuity whose derivatives will yield the delta function on the right-hand side.
4.3 Boundary Conditions–Maxwell’s Equations
As seen previously, boundary conditions for a field is embedded in the differential equation that the field satisfies. Hence, boundary conditions can be derived from the differential operator forms of Maxwell’s equations. In most textbooks, boundary conditions are obtained by integrating Maxwell’s equations over a small pill box [31,32,47]. To derive these boundary conditions, we will take an unconventional view: namely to see what sources can induce jump conditions on the pertinent fields. Boundary conditions are needed at media interfaces, as well as across current or charge sheets. As shall be shown, each of the Maxwell’s equations induces a boundary condition at the interface between two media or two regions separated by surface sources. 44 Electromagnetic Field Theory
4.3.1 Faraday’s Law
Figure 4.2: This figure is for the derivation of boundary condition induced by Faraday’s law. A local coordinate system can be used to see the boundary condition more lucidly. Here, the normaln ˆ =y ˆ and the tangential component tˆ=x ˆ.
For this, we start with Faraday’s law, which implies that
∂B ∇ × E = − (4.3.1) ∂t The right-hand side of this equation is a derivative of a time-varying magnetic flux, and it is a finite quantity. One quick answer we could ask is that if the right-hand side of the above equation is everywhere finite, could there be any jump discontinuity on the field E on the left hand side? The answer is no. To see this quickly, one can project the tangential field component and normal field component to a local coordinate system. In other words, one can think of tˆ andn ˆ as the localx ˆ andy ˆ coordinates. Then writing the curl operator in this local coordinates, one gets
∂ ∂ ∇ × E = xˆ +y ˆ × (ˆxE +yE ˆ ) ∂x ∂y x y ∂ ∂ =z ˆ E − zˆ E (4.3.2) ∂x y ∂y x
In simplifying the above, we have used the distributive property of cross product, and eval- uating the cross product in cartesian coordinates. The cross product gives four terms, but only two of the four terms are non-zero as shown above. ∂ ∂ Since the right-hand side of (4.3.1) is finite, the above implies that ∂x Ey and ∂y Ex have to be finite. In order words, Ex is continuous in the y direction and Ey is continuous in the x direction. Since in the local coordinate system, Ex = Et, then Et is continuous across the boundary. The above implies that
E1t = E2t (4.3.3) Magnetostatics, Boundary Conditions, and Jump Conditions 45 or the tangential components of the electric field is continuous at the interface. To express this in a coordinate independent manner, we have
nˆ × E1 =n ˆ × E2 (4.3.4) wheren ˆ is the unit normal at the interface, andn ˆ × E always bring out the tangential component of a vector E (convince yourself).
4.3.2 Gauss’s Law for Electric Flux From this Gauss’s law, we have
∇ · D = % (4.3.5) where % is the volume charge density.
Figure 4.3: A figure showing the derivation of boundary condition for Gauss’s law. Again, a local coordinate system can be introduced for simplicity.
Expressing the above in local coordinates (x, y, z as shown in Figure 4.3, then
∂ ∂ ∂ ∇ · D = D + D + D = % (4.3.6) ∂x x ∂y y ∂z z The boundary condition for the electric flux can be found by singularity matching. If there is a surface layer charge at the interface, then the volume charge density must be infinitely large or singular; hence, it can be expressed in terms of a delta function, or % = %sδ(z) in local coordinates. By looking at the above expression, the only term that can produce a ∂ δ(z) is from ∂z Dz. In other words, Dz has a jump discontinuity at z = 0; the other terms do not. Then ∂ D = % δ(z) (4.3.7) ∂z z s 46 Electromagnetic Field Theory
Integrating the above from 0 − ∆ to 0 + ∆, we get
0+∆
Dz(z) = %s (4.3.8) 0−∆ or in the limit when ∆ → 0,
+ − Dz(0 ) − Dz(0 ) = %s (4.3.9)
+ − + − where 0 = lim∆→0 0 + ∆, and 0 = lim∆→0 0 − ∆. Since Dz(0 ) = D2n, Dz(0 ) = D1n, the above becomes
D2n − D1n = %s (4.3.10)
In other words, a charge sheet %s can give rise to a jump discontinuity in the normal component of the electric flux D. Expressed in coordinate independent form, it is
nˆ · (D2 − D1) = %s (4.3.11)
Using the physical notion that an electric charge has electric flux D exuding from it, Figure 4.4 shows an intuitive sketch as to why a charge sheet gives rise to a discontinuous normal component of the electric flux D.
Figure 4.4: A figure intuitively showing why a sheet of charge gives rise to a jump discontinuiy in the normal component of the electric flux D.
4.3.3 Ampere’s Law Ampere’s law, or the generalized one, stipulates that ∂D ∇ × H = J + (4.3.12) ∂t Again if the right-hand side is everywhere finite, then H is a continuous field everywhere. However, if the right-hand side has a delta function singularity, due to a current sheet in 3D ∂D space, and that ∂t is regular or finite everywhere, then the only place where the singularity can be matched on the left-hand side is from the derivative of the magnetic field H or ∇ × H. In a word, H is not continuous. For instance, we can project the above equation onto a local coordinates just as we did for Faraday’s law. Magnetostatics, Boundary Conditions, and Jump Conditions 47
Figure 4.5: A figure showing the derivation of boundary condition for Ampere’s law. A local coordinate system is used for simplicity.
To be general, we also include the presence of a current sheet at the interface. A current sheet, or a surface current density becomes a delta function singularity when expressed as a volume current density; Thus, rewriting (4.3.12) in a local coordinate system, assuming that 4 J =xJ ˆ sxδ(z), then singularity matching, ∂ ∂ ∇ × H =x ˆ H − H =xJ ˆ δ(z) (4.3.13) ∂y z ∂z y sx The displacement current term on the right-hand side is ignored since it is regular or finite, and will not induce a jump discontinuity on the field; hence, we have the form of the right- hand side of the above equation. From the above, the only term that can produce a δ(z) ∂ singularity on the left-hand side is the − ∂z Hy term. Therefore, by singularity matching, we conclude that ∂ − H = J δ(z) (4.3.14) ∂z y sx
In other words, Hy has to have a jump discontinuity at the interface where the current sheet resides. Or that
+ − Hy(z = 0 ) − Hy(z = 0 ) = −Jsx (4.3.15) The above implies that
H2y − H1y = −Jsx (4.3.16)
But Hy is just the tangential component of the H field. In a word, the current sheet Jsx induces a jump discontinuity on the y component of the magnetic field. Now if we repeat the same exercise with a current with a y component, or J =yJ ˆ syδ(z), at the interface, we have
H2x − H1x = Jsy (4.3.17)
4The form of this equation can be checked by dimensional analysis. Here, J has the unit of A m−2, δ(z) −1 −1 has unit of m , and Jsx, a current sheet density, has unit of A m . 48 Electromagnetic Field Theory
Now, (4.3.16) and (4.3.17) can be rewritten using a cross product as
zˆ × (ˆyH2y − yHˆ 1y) =xJ ˆ sx (4.3.18)
zˆ × (ˆxH2x − xHˆ 1x) =yJ ˆ sy (4.3.19)
The above two equations can be combined as one, written in a coordinate independent form, to give
nˆ × (H2 − H1) = Js (4.3.20) where in this case here,n ˆ =z ˆ. In other words, a current sheet Js can give rise to a jump discontinuity in the tangential components of the magnetic field,n ˆ × H. This is illustrated intuitively in Figure 4.6.
Figure 4.6: A figure intuitively showing that with the understanding of how a single line current source generates a magnetic field (right), a cluster of them forming a current sheet will generate a jump discontinuity in the tangential component of the magnetic field H (left).
4.3.4 Gauss’s Law for Magnetic Flux Similarly, from Gauss’s law for magnetic flux, or that
∇ · B = 0 (4.3.21) one deduces that
nˆ · (B2 − B1) = 0 (4.3.22) or that the normal magnetic fluxes are continuous at an interface. In other words, since mag- netic charges do not exist, the normal component of the magnetic flux has to be continuous. The take-home message here is that the boundary conditions are buried in the differential operators. If there singular terms in Maxwell’s equations, then via the differential operators, the boundary conditions can be deduced. These boundary conditions are also known as jump condition if a current or a source sheet is present. Lecture 5
Biot-Savart law, Conductive Media Interface, Instantaneous Poynting’s Theorem
Biot-Savart law, like Ampere’s law was experimentally determined in around 1820 and it is discussed in a number of textbooks [31, 32, 48]. This is the cumulative work of Ampere, Oersted, Biot, and Savart. At this stage of the course, one has the mathematical tool to derive this law from Ampere’s law and Gauss’s law for magnetostatics. In addition, we will study the boundary conditions at conductive media interfaces, and introduce the instantaneous Poynting’s theorem.
49 50 Electromagnetic Field Theory 5.1 Derivation of Biot-Savart Law
Figure 5.1: A current element used to illustrate the derivation of Biot-Savart law. The current element generates a magnetic field due to Ampere’s law in the static limit. This law was established experimentally, but here, we will derive this law based on our mathematical knowledge so far.
From Gauss’ law and Ampere’s law in the static limit, and using the definition of the Green’s function, we have derived that ¦ µ J(r0) A(r) = dV 0 (5.1.1) 4π V R where R = |r − r0|. When the current element is small, and is carried by a wire of cross sectional area ∆a as shown in Figure 5.1, we can approximate the integrand as
J(r0)dV 0 ≈ J(r0)∆V 0 = (∆a)∆l ˆlI/∆a (5.1.2) | {z } | {z } ∆V J(r0)
In the above, ∆V = (∆a)∆l and ˆlI/∆a = J(r0) since J has the unit of amperes/m2. Here, ˆl is a unit vector pointing in the direction of the current flow. Hence, we can let the current element be
J(r0)∆V 0 = I∆l0 (5.1.3) where the vector ∆l0 = ∆lˆl, and 0 indicates that it is located at r0. Therefore, the incremental vector potential due to an incremental current element J(r0)∆V 0 is
µ J(r0)∆V 0 µ I∆l0 ∆A(r) = = (5.1.4) 4π R 4π R Biot-Savart law, Conductive Media Interface, Instantaneous Poynting’s Theorem 51
Since B = ∇ × A, we derive that the incremental B flux, ∆B due to the incremental current I∆l0 is
µI ∆l0 −µI 1 ∆B = ∇ × ∆A(r) = ∇ × = ∆l0 × ∇ (5.1.5) 4π R 4π R where we have made use of the fact that ∇×af(r) = −a×∇f(r) when a is a constant vector. The above can be simplified further making use of the fact that1
1 1 ∇ = − Rˆ (5.1.6) R R2 where Rˆ is a unit vector pointing in the r − r0 direction. We have also made use of the fact that R = p(x − x0)2 + (y − y0)2 + (z − z0)2. Consequently, assuming that the incremental length becomes infinitesimally small, or ∆l → dl, we have, after using (5.1.6) in (5.1.5), that
µI 1 dB = dl0 × Rˆ 4π R2 µIdl0 × Rˆ = (5.1.7) 4πR2
Since B = µH, we have
Idl0 × Rˆ dH = (5.1.8) 4πR2 or
¢ I(r0)dl0 × Rˆ H(r) = (5.1.9) 4πR2 which is Biot-Savart law, first determined experimentally, now derived using electromagnetic field theory.
1This is best done by expressing the ∇ operator in spherical coordinates. 52 Electromagnetic Field Theory 5.2 Shielding by Conductive Media 5.2.1 Boundary Conditions–Conductive Media Case
Figure 5.2: The schematics for deriving the boundary condition for the current density J at the interface of two conductive media.
In a conductive medium, J = σE, which is just a statement of Ohm’s law. From the current continuity equation, which is derivable from Ampere’s law and Gauss’ law for electric flux, one gets ∂% ∇ · J = − (5.2.1) ∂t If the right-hand side is everywhere finite, it will not induce a jump discontinuity in the current. Moreover, it is zero for static limit. Hence, just like the Gauss’s law case, the above implies that the normal component of the current Jn is continuous, or that J1n = J2n in the static limit. In other words,
nˆ · (J2 − J1) = 0 (5.2.2) Hence, using J = σE, we have
σ2E2n − σ1E1n = 0 (5.2.3)
The above has to be always true in the static limit irrespective of the values of σ1 and σ2. But Gauss’s law implies the boundary condition that
ε2E2n − ε1E1n = %s (5.2.4)
The above equation is incompatible with (5.2.3) unless %s 6= 0. Hence, surface charge density or charge accumulation is necessary at the interface, unless σ2/σ1 = ε2/ε1. This is found in semiconductor materials which are both conductive and having a permitivity: interfacial charges appear at the interface of two semi-conductor materials. Biot-Savart law, Conductive Media Interface, Instantaneous Poynting’s Theorem 53
5.2.2 Electric Field Inside a Conductor The electric field inside a perfect electric conductor (PEC) has to be zero by the explanation as follows. If medium 1 is a perfect electric conductor, then σ → ∞ but J1 = σE1. An infinitesimal E1 will give rise to an infinite current J1. To avoid this ludicrous situation, E1 has to be 0. This implies that D1 = 0 as well.
Figure 5.3: The behavior of the electric field and electric flux at the interface of a perfect electric conductor and free space (or air).
Since tangential E is continuous, from Faraday’s law, it is still true that
E2t = E1t = 0 (5.2.5) orn ˆ × E = 0. But since
nˆ · (D2 − D1) = %s (5.2.6) and that D1 = 0, then
nˆ · D2 = %s (5.2.7)
So surface charge density has to be nonzero at a PEC/air interface. Moreover, normal D2 6= 0, tangential E2 = 0: Thus the E and D have to be normal to the PEC surface. The sketch of the electric field in the vicinity of a perfect electric conducting (PEC) surface is shown in Figure 5.3. The above argument for zero electric field inside a perfect conductor is true for electro- dynamic problems. However, one does not need the above argument regarding the shielding of the static electric field from a conducting region. In the situation of the two conducting 54 Electromagnetic Field Theory objects example below, as long as the electric fields are non-zero in the objects, currents will keep flowing. They will flow until the charges in the two objects orient themselves so that electric current cannot flow anymore. This happens when the charges produce internal fields that cancel each other giving rise to zero field inside the two objects. Faraday’s law still applies which means that tangental E field has to be continuous. Therefore, the boundary condition that the fields have to be normal to the conducting object surface is still true for elecrostatics. A sketch of the electric field between two conducting spheres is show in Figure 5.4.
Figure 5.4: The behavior of the electric field and flux outside two conductors in the static limit. The two conductors need not be PEC, and yet, the fields are normal to the interface.
5.2.3 Magnetic Field Inside a Conductor
We have seen that for a finite conductor, as long as σ 6= 0, the charges will re-orient themselves until the electric field is expelled from the conductor; otherwise, the current will keep flowing until E = 0 or ∂tE = 0. In a word, static E is zero inside a conductor. But there are no magnetic charges nor magnetic conductors in this world. Thus this physical phenomenon does not happen for magnetic field: In other words, static magnetic field cannot be expelled from an electric conductor. However, a magnetic field can be expelled from a perfect conductor or a superconductor. You can only fully understand this physical phenomenon if we study the time-varying form of Maxwell’s equations. In a perfect conductor where σ → ∞, it is unstable for the magnetic field B to be nonzero. As time varying magnetic field gives rise to an electric field by the time-varying form of Faraday’s law, a small time variation of the B field will give rise to infinite current flow in a perfect conductor. Therefore to avoid this ludicrous situation, and to be stable, B = 0 in a perfect conductor or a superconductor. Biot-Savart law, Conductive Media Interface, Instantaneous Poynting’s Theorem 55
So if medium 1 is a perfect electric conductor (PEC), then B1 = H1 = 0. The boundary condition for magnetic field from Ampere’s law
nˆ × (H2 − H1) =n ˆ × H2 = Js (5.2.8)
which is the jump condition for the magnetic field. In other words, a surface current Js has to flow at the surface of a PEC in order to support the jump discontinuity in the tangential component of the magnetic field. From Gauss’s law,n ˆ · B is always continuous at an interface because of the absence of magnetic charges. The magnetic flux B1 is expelled from the perfect conductor makingn ˆ · B1 = 0 zero both inside and outside the PEC (due to the continuity of normal B at an interface). And hence, there is no normal component of the B1 field at the interface. Therefore, the boundary condition for B2 becomes, for a PEC,
nˆ · B2 = 0 (5.2.9)
The B field in the vicinity of a perfect conductor surface is as shown in Figure 5.5.
Figure 5.5: Sketch of the magnetic flux B around a perfect electric conductor. It is seen that nˆ · B = 0 at the surface of the perfect conductor.
When a superconductor cube is placed next to a static magnetic field near a permanent magnet, eddy current will be induced on the superconductor. The eddy current will expel the static magnetic field from the permanent magnet, or in a word, it will produce a magnetic dipole on the superconducting cube that repels the static magnetic field. Since these two magnetic dipoles are of opposite polarity, they repel each other, and cause the superconducting cube to levitate on the static magnetic field as shown in Figure 5.6. 56 Electromagnetic Field Theory
Figure 5.6: Levitation of a superconducting disk on top of a static magnetic field due to expulsion of the magnetic field from the superconductor.. This is also known as the Meissner effect (figure courtesy of Wikimedia).
5.3 Instantaneous Poynting’s Theorem
Before we proceed further with studying energy and power, it is habitual to add fictitious magnetic current M and fictitious magnetic charge %m to Maxwell’s equations to make them symmetric mathematically.2 To this end, we have
∂B ∇ × E = − − M (5.3.1) ∂t ∂D ∇ × H = + J (5.3.2) ∂t ∇ · D = % (5.3.3)
∇ · B = %m (5.3.4)
Consider the first two of Maxwell’s equations where fictitious magnetic current is included and that the medium is isotropic such that B = µH and D = εE. Next, we need to consider only the first two equations (since in electrodynamics, by invoking charge conservation, the third and the fourth equations are derivable from the first two). They are
∂B ∂H ∇ × E = − − M = −µ − M (5.3.5) ∂t i ∂t i ∂D ∂E ∇ × H = + J = ε + J + σE (5.3.6) ∂t ∂t i where Mi and Ji are impressed current sources. They are sources that are impressed into the system, and they cannot be changed by their interaction with the environment [50]. Also, for a conductive medium, a conduction current or induced current flows in addition to impressed current. Here, J = σE is the induced current source in the conductor. Moreover, J = σE is similar to ohm’s law. We can show from By dot multiplying (5.3.5) with H, and
2Even though magnetic current does not exist, electric current can be engineered to look like magnetic current as shall be learnt. Biot-Savart law, Conductive Media Interface, Instantaneous Poynting’s Theorem 57 dot multiplying (5.3.6) with E, we can show that ∂H H · ∇ × E = −µH · − H · M (5.3.7) ∂t i ∂E E · ∇ × H = εE · + E · J + σE · E (5.3.8) ∂t i Using the identity, which is the same as the product rule for derivatives, we have3
∇ · (E × H) = H · (∇ × E) − E · (∇ × H) (5.3.9)
Therefore, from (5.3.7), (5.3.8), and (5.3.9) we have
∂H ∂E ∇ · (E × H) = − µH · + εE · + σE · E + H · M + E · J (5.3.10) ∂t ∂t i i
To elucidate the physical meaning of the above, we first consider σ = 0, and Mi = Ji = 0, or the absence of conductive loss and the impressed current sources. Then the above becomes ∂H ∂E ∇ · (E × H) = − µH · + εE · (5.3.11) ∂t ∂t Rewriting each term on the right-hand side of the above, we have ∂H 1 ∂ ∂ 1 ∂ µH · = µ H · H = µ|H|2 = W (5.3.12) ∂t 2 ∂t ∂t 2 ∂t m ∂E 1 ∂ ∂ 1 ∂ εE · = ε E · E = ε|E|2 = W (5.3.13) ∂t 2 ∂t ∂t 2 ∂t e where |H(r, t)|2 = H(r, t) · H(r, t), and |E(r, t)|2 = E(r, t) · E(r, t). Then (5.3.11) becomes ∂ ∇ · (E × H) = − (W + W ) (5.3.14) ∂t m e where 1 1 W = µ|H|2,W = ε|E|2 (5.3.15) m 2 e 2 Equation (5.3.14) is reminiscent of the current continuity equation, namely that, ∂% ∇ · J = − (5.3.16) ∂t which is a statement of charge conservation. In other words, time variation of current density at a point is due to charge density flow into or out of the point. The vector quantity
Sp = E × H (5.3.17)
3The cyclical identity that a · (b × c) = c · (a × b) = b · (c × a) is useful for the derivation. 58 Electromagnetic Field Theory is called the Poynting’s vector , and (5.3.14) becomes
∂ ∇ · S = − W (5.3.18) p ∂t t where Wt = We + Wm is the total energy density stored in the electric and magnetic fields. The above is similar to the current continuity equation in physical interpretation mentioned above. Analogous to that current density is charge density flow, power density is energy density flow. It is easy to show that Sp has a dimension of watts per meter square, or power density, and that Wt has a dimension of joules per meter cube, or energy density. Now, if we let σ 6= 0, then the term to be included is then σE · E = σ|E|2 which has the unit of S m−1 times V2 m−2, or W m−3 where S is siemens. We arrive at this unit by 1 V 2 noticing that 2 R is the power dissipated in a resistor of R ohms with a unit of watts. The reciprocal unit of ohms, which used to be called mhos is now called siemens. With σ 6= 0, (5.3.18) becomes
∂ ∂ ∇ · S = − W − σ|E|2 = − W − P (5.3.19) p ∂t t ∂t t d
2 Here, ∇·Sp has physical meaning of power density oozing out from a point, and −Pd = −σ|E| has the physical meaning of power density dissipated (siphoned) at a point by the conductive loss in the medium which is proportional to −σ|E|2.
Now if we set Ji and Mi to be nonzero, (5.3.19) is augmented by the last two terms in (5.3.10), or
∂ ∇ · S = − W − P − H · M − E · J (5.3.20) p ∂t t d i i
The last two terms can be interpreted as the power density supplied by the impressed currents Mi and Ji. Therefore, (5.3.20) becomes
∂ ∇ · S = − W − P + P (5.3.21) p ∂t t d s where
Ps = −H · Mi − E · Ji (5.3.22)
Here, Ps is the power supplied by the impressed current sources. These terms are positive if H and Mi have opposite signs, or if E and Ji have opposite signs. The last terms reminds us of what happens in a negative resistance device or a battery.4 In a battery, positive charges move from a region of lower potential to a region of higher potential (see Figure 5.7). The positive charges move from one end of a battery to the other end of the battery. Hence, they are doing an “uphill climb” driven by chemical processes within the battery.
4A negative resistance has been made by Leo Esaki [51], winning him a share in the Nobel prize. Biot-Savart law, Conductive Media Interface, Instantaneous Poynting’s Theorem 59
Figure 5.7: Figure showing the dissipation of energy as the current flows around a loop. A battery can be viewed as having a negative resistance.
−3 In the above, one can easily work out that Ps has the unit of W m which is power supplied density. One can also choose to rewrite (5.3.21) in integral form by integrating it over a volume V and invoking the divergence theorem yielding ¢ ¢ ¢ ¢ d dS · Sp = − WtdV − PddV + PsdV (5.3.23) S dt V V V The left-hand side is ¢ ¢
dS · Sp = dS · (E × H) (5.3.24) S S which represents the power flowing out of the surface S. 60 Electromagnetic Field Theory Lecture 6
Time-Harmonic Fields, Complex Power
The analysis of Maxwell’s equations can be greatly simplified by assuming the fields to be time harmonic, or sinusoidal (cosinusoidal). Electrical engineers use a method called phasor technique [32,52], to simplify equations involving time-harmonic signals. This is also a poor- man’s Fourier transform [53]. That is one begets the benefits of Fourier transform technique without knowledge of Fourier transform. Since only one time-harmonic frequency is involved, this is also called frequency domain analysis.1 Phasors are represented in complex numbers. From this we will also discuss the concept of complex power.
1It is simple only for linear systems: for nonlinear systems, such analysis can be quite unwieldy. But rest assured, as we will not discuss nonlinear systems in this course.
61 62 Electromagnetic Field Theory 6.1 Time-Harmonic Fields—Linear Systems
Figure 6.1: A commemorative stamp showing the contribution of Euler (courtesy of Wikipedia and Pinterest).
To learn phasor technique, one makes use the formula due to Euler (1707–1783) (Wikipedia) ejα = cos α + j sin α (6.1.1) √ where j = −1 is an imaginary number.2 From Euler’s formula one gets cos α =